April 05, 2006

How's that for reactivity

A bunch of us updated our versions of Ubuntu this afternoon. update-notifier 0.41.11 had a problem that caused it to crash when Gnome started.

Three hours later the problem was fixed.

My goodness do I feel bad about all of my old bugs now. Maybe I need to get us into a mode like this. Emotionally I'm not a very just-in-time guy, but there are some things to be said for operating like that.

Posted by Mark at 08:17 PM | TrackBack

April 04, 2006

File sharing and legal trouble

BBC News online has an article about how music industry middlemen are fighting to sue people for sharing tunes they bought (or got from somebody else). The problem is apparently big enough to eat into sales and profits.

What's file sharing going to be like a few years from now? Will home bandwidth increase by a factor of 10 over the next 10 years? If so, the legislation is going to take us in the direction of drug legislation now, except that children will be sharing movies, music, and video games rather than getting high.

I can imagine the police knocking on my door to lead me away to jail because one of the three kids downloaded a cracked version of some recently released movie. That said, maybe I'll be unemployable by then and won't even be able to afford normal wireless Internet access.

Posted by Mark at 08:45 PM | TrackBack

March 29, 2006

Subversion

My brother Matt set up a subversion repository on this site. It seems to work fine. If you want to add more date formats, go ahead. (You'll have to ask one of us before you can commit, I guess.)

The nice thing about subversion is that if you know cvs, getting started is like falling off a log. Furthermore as (the other) Matt explained, each commit is a revision of the whole thing. You proceed along the trunk, and have another directory hierarchy for branches, and yet another for tags. I guess that's the suggested default setup. Whatever it is, it's a clear product management default that makes you think about the project in the right way from the start.

Not that we're going to be releasing any earth shattering software from that repository, but it's nice to know that given enough monkeys and enough keyboards, we have the right technology to do so. Hey, the Internet is a big place.

Posted by Mark at 08:57 PM | TrackBack

March 28, 2006

Tail wagging dog

Ludo pointed me a few days ago to Dan Bricklin's article, When The Long Tail Wags The Dog. Dan looks at how the overall value of general purpose tools such as the spreadsheet or the telephone can outstrip that of special purpose tools, in particular when there exist highly personal, "must have" applications the general purpose tools can enable. The general purpose tool, "is of high value for a wider range," than its special purpose counterpart.

In agreeing with the general thesis, I'd underline that the general purpose tool has to be more like a telephone than a Swiss army knife. Dan alludes to this when he writes, "A general purpose tool that most people can't use easily or understand how to apply in new situations will be less popular." I know a number of writers who refuse to learn vi, and it's probably because gedit or whatever it is they use is more like a telephone than a Swiss army knife.

Posted by Mark at 08:41 PM | TrackBack

March 27, 2006

Ugly ducklings officially okay

Jared Spool wrote about how what he sees as ugly sites can still attract users.

At some point, (and I don’t quite know where that point is at this time,) fashion, visual appeal, and aesthetic comfort becomes a priority to the audience. At that point, you better be ready or else you’ll look dated and amateurish. But get there too early and you’re wasting valuable resources on something users don’t care about.

Hmm. Function over form. Isn't he just pandering to those of us aesthetically challenged folks who only wonder whether Mac OS X would be worth it when trying to edit home videos?

Interestingly I notice that an iPod comparable to the MP3 player I bought is currently only $10 more expensive at Amazon.com. I'd pay $10 more for the same computer if it made video editing easy (as long as it's still UNIX-like underneath).

Posted by Mark at 07:52 PM | Comments (2) | TrackBack

March 23, 2006

Too complicated

Tim was asking me this evening how to copy some text from a Wikipedia article on Star Wars to another file. Even that I could see was not sinking in instantly.

Then later I came up here and had network problems. I could get an IP address from the router, and the router was getting two name server IP addresses from the DHCP server it gets the address from. But lookups weren't working. I finally found a printed message from my ISP in which two other, different DNS IP addresses were mentioned. After adding those to /etc/resolv.conf I could browse the web. No sweat.

Then it dawned on me that all of this is too complicated, that I'm warped by my job into thinking that fixing the DNS server IP addresses in /etc/resolv.conf by hand is a normal sort of thing to have to do on the client side. We need to figure out how to make this stuff simpler not just to develop to, but for normal people to debug.

Posted by Mark at 09:29 PM | TrackBack

March 21, 2006

Webcam on Ubuntu, part VI

ekiga-20060321.jpg On a tip from Christian Juner, I installed Ekiga on Breezy. It seems to work on Dapper Drake. It also seems to work, almost, on Breezy.

For a reason I've not yet managed to discover, my registration is coming back "Forbidden." Yet I can login at ekiga.net. The log, Tools > General History, is concise. The folks on the mailing list who seem to have had the same symptoms all got around it. One person said a change of passwords worked.

I've not yet tried to learn which ports I might be supposed to open. Tried to interest Luke in trying this out at work since he was on Windows. He saw the intructions for using SIP with whatever Microsoft's comparable software is called and gave up before trying.

Posted by Mark at 09:17 PM | TrackBack

Everything slips

Yet another big software project schedule has slipped. Ubuntu Dapper Drake is slipping to June 1.

The good thing is that it's already fairly solid, if not so stable. For my use it already seems to work.

Curiously the documentation freeze is coming before the user interface freeze. Can Ubuntu docsters count on the specs so completely they're ready to finish the doc before the developers quit tweaking? I need to get my head around that one.

Posted by Mark at 06:21 AM | TrackBack

March 16, 2006

apt-get dist-upgrade?

On the laptop where I write this entry, I used apt-get dist-upgrade (or did I do that in Synaptic?) to move from Ubuntu 6.04 Flight CD 4 to Flight CD 5. This system would be awful to use in real production. The problem is not the stability or the bugs. It works fine and I haven't run into any bugs. The problem is the volume of updates. Over 100 megabytes this morning. Another 100 megabytes this evening if I had the energy to get the updates.

But should I do apt-get dist-upgrade on the PC where we have all our stuff? I know some of the apps have been dropped, and some have changed.

Matt did the same sort of thing with yum the other day and found he had to redo his network configuration afterward. That's all he found immediately, but I didn't ask him what he found next.

Sometimes we think of Solaris as being a bit stodgy, but at least you can upgrade most of the time without fear that your whole configuration will be hosed.

Posted by Mark at 09:44 PM | Comments (1) | TrackBack

Building playlists

A while ago I ripped my CD collection (except for a couple of recalcitrant CDs that wouldn't rip properly for one reason or another). Maybe when the children have grown I'll be able to sit in the living room and listen to a CD, but for now I need them on the MP3 player.

Listening while jogging along the road is not so good, but listening while sitting in a running car is worse. That's probably why so many adults listen to talk radio. Classical music in my car is like drinking from a mud puddle.

But sometimes I'd like to listen while I'm at the computer. In that case, I'd like to have album playlists. And I don't want to do that by hand for all the CDs I ripped. Do I write my own m3u file builder? Can I find a good one online?

Posted by Mark at 09:33 PM | TrackBack

March 14, 2006

15 cents/(GB*month), plus 20 cents/GB uploads

I noticed on TechCrunch an entry about Amazon's Simple Storage Service. It looks like they've put a price on easily extensible online storage.

What they don't seem to have priced yet is web storage that is guaranteed not to disappear without notice. Way down in the disclaimers section of their agreement you find that they're not going to be liable for any disasters.

Posted by Mark at 09:38 PM | TrackBack

March 12, 2006

Device recognition better now

Tim wanted me to put Windows 98 back on the Vaio instead of Ubuntu. He wants to use old games that I cannot get to run on Ubuntu with Wine, the Windows emulator.

First of all, the Vaio recovery CDs that reinstall Windows 98 do not include a utility to change the partition table. Good thing I had an old Gentoo CD lying around for fdisk.

Next, there are bad blocks on the disk. Windows 98 is going to run into them some day and break. Will try the disk management tools, but don't know if they'll work.

Finally, using the wireless card with Windows 98 is a pain. The setup and driver didn't fit on a floppy, so I had to burn a CD. It took me a while just to find the driver. Seems to work now. A disappointing bit is that this computer has one PCMCIA slot, so it's either the CD drive or the network, but not both.

Posted by Mark at 10:16 AM | TrackBack

March 04, 2006

What I cannot find online

I've been looking for evidence it'd be possible to take a shower after the semi-marathon Sunday, but cannot find anything. Nor is it clear there'd be an alternative, like showers at the Gare de Lyon for example.

Maybe I'll just have to stink all the way home tomorrow evening. Of course if the weather turns out to be as cool as predicted, we might not sweat that much after all.

Posted by Mark at 07:03 AM | TrackBack

February 27, 2006

Webcam on Ubuntu, part V

This evening both audio and video worked between GnomeMeeting and NetMeeting. I'd turned off sound events at startup in Ubuntu. That seems to have the effect that the audio settings in GnomeMeeting work, rather than breaking silently.

20060217.jpg

Thank you to the developers who put whatever is built in to break feedback loops.

Posted by Mark at 10:13 PM | TrackBack

February 26, 2006

Webcam on Ubuntu, part IV

I can manage to get GnomeMeeting from one system to call NetMeeting on the laptop from work, which has Windows on it. I can get audio from NetMeeting that plays back through GnomeMeeting and vice versa, but I cannot get audio with the webcam to work in GromeMeeting. I cannot even get the text to go through for reasons I do not understand. Quite frustrating.

Furthermore, it's not clear how I can post my H323 address out through the NAT, nor can I figure out how I'm going to be able to find someone else's H323 address. In some ways it feels like this stuff's not quite ready for prime time on Ubuntu.

Posted by Mark at 02:51 PM | Comments (2) | TrackBack

Webcam on Ubuntu, part III

Hmm. Hmm. The audio works fine at the command line. I've recorded a short test with the rec and play commands.

rec -d /dev/audio1 test.wav; play -d /dev/audio test.wav

The Gnome applications I've tried, such as gnome-sound-recorder, do not seem even to be using the microphone.

Posted by Mark at 09:58 AM | TrackBack

Webcam on Ubuntu, part II

Although I can load the quicktime module I built and can get the webcam to work, loading the module at system startup results in things going haywire. Here's what I see in dmesg:

[4296518.087000] quickcam [52.080121]: failed qc_capt_get()=-90
[4296518.087000] quickcam [52.080128]: failed qc_v4l_read()=-90
[4296518.111000] quickcam [52.104330]: submit ISOC_URB 0 failed
[4296518.111000] quickcam [52.104344]: failed qc_isoc_init()=-90
[4296518.111000] quickcam: unable start isoc

When I remove and reinsert the module:

[4296548.265000] quickcam [22.253775]: ----------LOADING QUICKCAM MODULE------------
[4296548.265000] quickcam [22.253787]: struct quickcam size: 3904
[4296548.270000] quickcam: QuickCam USB camera found (driver version QuickCam Messenger/Communicate USB $Date: 2004/12/30 10:00:00 $)
[4296548.270000] quickcam: Kernel:2.6.12-10-386 bus:1 class:FF subclass:FF vendor:046D product:08F6
[4296548.270000] quickcam [22.258686]: poisoning qc in qc_usb_init
[4296548.280000] quickcam [22.268615]: E00A contains 08F6
[4296548.280000] quickcam: Sensor VV6450 detected
[4296548.326000] quickcam [22.315156]: Quickcam snapshot button registered on usb-0000:00:02.0-2/input0
[4296548.362000] quickcam: Registered device: /dev/video0
[4296548.362000] usbcore: registered new driver quickcam

Hmm. At least /dev/video0 works after that. What must I do to get the microphone to work?

Posted by Mark at 08:51 AM | TrackBack

February 25, 2006

Webcam on Ubuntu

webcam-20050225.jpg You can tell by my expression that getting everything working is not quite easy, yet.

First, I followed instructions from somebody in Belgium. A nice scrip, though I didn't know that lsusb would tell me what I needed to change in the Perl line. (This QuickCam Messenger is product id 08f6.) Also, the latest version of the driver is 1.1 rather than 0.8.

Next, I needed to get gcc-3.4 installed. The one by default is 4.0.

After a bit of playing around, I can see the video feed. Yet I cannot get the audio worked out. Will have another look later.

Posted by Mark at 03:37 PM | Comments (2) | TrackBack

February 23, 2006

Web-based web page editor

Most folks creating their own web pages are probably already doing it through their browser, by blogging. There are however cases where you may want to create a static page or two. We create them at work, for example, as cover pages for documentation snapshots we post for review, for some of our plans, pages we expect people to link to directly and perhaps consult frequently.

If you know HTML, you probably end up editing your static pages in something like vim or bluefish. If you don't know HTML, maybe you export to HTML from a office program like OOo Writer or OOo Calc. A recent alternative for folks who do static pages and want to be able to edit them without having installed software is Google's beta Page Creator.

20060223.png I gave Page Creator a whirl. It puts your home page under a URL at their site. You can see the page I created here.

If you wanted to create static sites with this method, you could use HTML redirects from your main site, or just link from your blog. It's not immediately clear how you're supposed to do a backup. Maybe the idea is that you crawl through the pages starting from your home page. Maybe in this Web 2.0 world, you leave the backup to someone else.

Posted by Mark at 09:27 AM | TrackBack

February 20, 2006

Paper for digital storage

Today I went to see my doctor to finish off the high blood pressure question. He looked at the results from a couple of weeks ago when I wore the portable blood pressure measurement device all night, and spent a little more time than the cardiologist saying the high readings appear to have been a fluke. Today he said my blood pressure was douze sept. I guess he meant 12.7 over something not worth mentioning. Either that or he mean 12/7, which is 120/70. Neither reading is cause for alarm.

The interesting part of the visit came when he confused me, or at least my job, with another American he treats. This other guy apparently is moving to China to work on productizing the use of paper for digital storage. He wasn't talking about punched tape, but instead storing digital information in paper fibers. Presumably we have better preservation technology for paper than we do for magnetic or optical media. Perhaps we now have techniques for cramming lots more onto paper.

I recall reading something this, but cannot recall where, nor what would bring it back up. Google's not helping.

Posted by Mark at 06:05 PM | Comments (3) | TrackBack

February 19, 2006

Favorite option

20060219.png

The 3D acceleration seemed slow on this system. This is the test.

$ glxgears -iacknowledgethatthistoolisnotabenchmark
6004 frames in 5.0 seconds = 1200.728 FPS
7212 frames in 5.0 seconds = 1442.209 FPS
7209 frames in 5.0 seconds = 1441.750 FPS
7213 frames in 5.0 seconds = 1442.500 FPS
7060 frames in 5.0 seconds = 1411.919 FPS
6305 frames in 5.0 seconds = 1256.035 FPS
X connection to :0.0 broken (explicit kill or server shutdown).

You can also get -iacknowledgethatthistoolisnotabenchmark with -printfps. Too bad.

Posted by Mark at 08:54 PM | TrackBack

More online robbery

Now that I'm looking into things at Free.fr, I also see that despite their posted tarifs for telephone calls, they've been charging us for calls they say are no additional cost.

It's not much money, but needs to be corrected. Since it's going to cost me time to fix their mess, I wonder what I can do to cost them as much as possible. Maybe some letters to editors would help.

Update: Aha! The thieves seem to have protected their theft on a technicality. They don't propose the offer, but I have to go read 20 pages of fine print again to find the differences between what they proposed to me at the end of December and what I'd have to accept now that I'm signed up. Hmm. So if I'm going to get a fair shake for my work, it's going to need to cost them dearly.

Another update: It's not 20 pages. It's 41.

Posted by Mark at 01:03 PM | TrackBack

Connection down all morning

I have not yet figured out where the problem occurred, but our DSL connection was interrupted again all morning. Our provider of course provides a procedure that systematically blames France Telecom. Yet I don't recall ever having more than momentary outtages, always fixed by restarting the software connection, when France Telecom was our provider.

The technical people at our provider are probably aware this sort of thing happens, because in general they're quite competent. Maybe it's at the transfer of the ATM frames from FT equipment to Free.fr equipment. Maybe it's just that the bankers who own Free.fr figure consumers will have no recourse, and will just blame FT, even though they're aware the problem is on Free.fr's side. If we, Free.fr's customers, could find that out, we could perhaps force them either to give us money back pro rata, or with a penalty, such that they'd be forced to provide higher quality service.

Posted by Mark at 12:48 PM | TrackBack

February 14, 2006

And you thought my running entries were too much

This guy uses his watch plus Google Maps to upload traces of his actual runs to his blog.

He's willing to share the software, but you need a particular model of watch for it to work. Not sure I'd be willing to make the investment. I'm waiting for the heartbeat monitor to that sends email and lets you keep a running verbal account of how you feel, altitude, temperature, blood concentrations, brain wave activity, shoe wear and tear, etc., and uploads all that via wireless into an online expert system coaching program that whispers advice through your headphones.

Posted by Mark at 08:47 PM | Comments (3) | TrackBack

Reintermediation

According to MSNBC, Warner Music experienced a large relative uptick in digital music sales this past quarter.

Warner Music, which went public last May, said digital revenues surged by 176 per cent to $69m in the first quarter, from the $25m achieved the previous year and up 30 per cent from the previous quarter. Digital sales now account for 7 per cent of overall revenue, the company said.

To what extent does that offset the money the majors claim to have lost by being disintermediated from the physical music media markets?

Posted by Mark at 08:38 PM | TrackBack

February 10, 2006

Broken Gnome, part II

By the way, I found out Gnome was breaking because I'd set LD_LIBRARY_PATH in my environment, and some incompatible, contributed software libraries got linked in first. Boom.

The thing I hadn't see was that you can start a failsafe session on the SunRay, which doesn't try to run gnome-session. Good for debugging. Sort of.

After I removed LD_LIBRARY_PATH from my environment, I had to figure out which apps wouldn't open without it. And I still get two instances of gconfd-2. Shutting down the session only kills one.

Maybe I'm not supposed to have the contributed software binaries in my PATH?

Posted by Mark at 09:37 PM | TrackBack

February 06, 2006

Fat browser

Somebody's blog entry on Understanding memory usage on Linux got Slashdotted. I went over to look. It's mainly about the effect of shared library code and looking at this with pmap -d.

The example was with a KDE text editor, which apparently uses lots of shared library code, and so looks much more bloated under ps than it actually is. There are however other applications that not only look bloated, but also are pretty big.

$ ps -e | grep firefox
10447 ? 00:07:38 firefox-bin
$ pmap -d 10447 | tail -1
mapped: 156204K writeable/private: 118824K shared: 768K

That's Firefox with the aforementioned article opened, nothing else.

Posted by Mark at 10:16 PM | TrackBack

February 03, 2006

movein.sh ?

My brother Matt and I were thinking about online backups. If you're like me, your ISP provides more space than you need to backup everything in your family's $HOMEs. The biggest pain in reinstalling is transferring files and setting up the $HOMEs again including configuration of preferences, etc. At home I don't want to run a file server for /home in the basement all the time.

So Matt's suggesting a movein.sh script that gets everything and sets it up for you, and you don't have to do that manually every time you bring a new machine online. Hmm. I wonder about authentification as well. What if I had LDAP-based naming on my machines, but left the server running at my ISP? It would be the opposite of high end.

In any case if a web-based service took the user auth and user data problems away from a particular home system, that might have a tendency to encourage people to change systems more often. I'd also be one step closer to work from anywhere.

Posted by Mark at 08:30 AM | Comments (3) | TrackBack

February 01, 2006

/home at home, part II

I've finally taken the time to move everything in our /home at home over to a single disk. I checked that lots of things still worked, then eliminated the duplicate files. This leaves me with space to capture and edit video, or try out other OSs. If only I had the energy and the time.

Posted by Mark at 09:50 PM | TrackBack

January 25, 2006

DSL up

Nothing I did made any difference. Nathalie said the problem disappeared during the morning.

I'm guessing it was my DSL provider, Free, who broke something and now probably isn't going to tell me about it.

Posted by Mark at 02:49 PM | TrackBack

January 23, 2006

Dapper Drake flight CD 3, part II

My first job in QA is volunteer work. Not exactly high value add stuff. I just went down the menus and tried to start all the apps. There were several that didn't work in this alpha version, so I logged bugs, a few of which ended up being duplicates.

It's a good way to learn a little bit about the packaging system, though mine attempts to find causes are mostly stabs in the dark. After one of the Ubuntu core dev guys, Sebastien, noticed a set of my bugs were all due to the same missing library for handling SVG images, I realized one of my others was probably just a missing dependency that didn't get flagged somehow. When I found the right package, the problem disappeared.

Somebody in marketing once talked about "unattended complexity." Rob and Luke had that on their whiteboard. Flagging dependencies brings those words to mind. We sometimes have people ask, "What's the minimum set of packages we need to install to run your software in production?" If you start doing anything very interesting, that becomes a hard question to answer.

Posted by Mark at 09:17 PM | TrackBack

January 17, 2006

whereami for laptop network configuration

One of the Ubuntu Wiki's sent me to whereami, the network autoconfiguration utility you can get for Debian and therefore Ubuntu.

Configuration is done using two files, detect.conf and whereami.conf, the former being used to set up your network connection including which interface you use and how you get that configured, the second for other actions you want to script depending on the network configuration. /etc/init.d/whereami start runs at boot time, though you can also of course sudo /etc/init.d/whereami stop; sudo /etc/init.d/whereami start any time you change networks.

So far the only thing I do in whereami.conf is set proxy preferences for Firefox, but since it's just a shell script with location information, I could do just about anything. Highly recommended if you have a portable computer.

Posted by Mark at 04:19 PM | Comments (1) | TrackBack

January 16, 2006

Merging $HOMEs

There's a nice Linux Journal article on unison, which is a good way to keep in sync once you were foresightful enough to set a home directory up properly.

But why didn't I think in the first place about setting up one of the PC disk partitions as a permanent $HOME for everybody's stuff? Now I'm trying to go back and synchronize things between the two systems I have installed on the PC, in preparation of syncing with the laptops, and eventually organizing a sync with my work $HOME for those mornings or afternoons when I'd get more done by shutting myself away and ploughing through some work.

I read a NYT editorial in which the writer was complaining about slow network access. Apparently some Scandinavian ISPs are offering Gigabit speeds to your house for the equivalent of about $100/mo. If I had that kind of access, the very next thing I'd want is all my data mounted from the network and backed up by competent professionals. Mobile high speed access and almost no local data. That would be the cat's meow.

Posted by Mark at 09:05 PM | TrackBack

January 14, 2006

Wireless access at home, part II

The old laptop I'm using to write this entry has a PCMCIA card plugged in, and so is the third wireless device configured to use the home network. I'm at Tim's desk.

My hope is that I can get him to play some of the educational games on the computer if he has network access, too. In the past I tried to get him to do that, but he figured a computer without an internet hookup is useless. So the only one he used was the PC upstairs, the main machine I want to use as well.

Although it's PIII 500 MHz processor and 128 MB RAM make this Vaio a relic, the only serious shortcoming I see against using this one for what Tim and Emma and potentially Diane want to do is that the sound card is somehow broken, probably as the result of a fall a while ago. No way to hear, "Hi! I'm Barbie!" when you visit Barbie.com.

Posted by Mark at 10:41 AM | TrackBack

January 13, 2006

New hardware breaks video device, part II

Jean-Luc and Fabio suggested the problem was with interrupts. The fix that worked was to boot with pci=noacpi. So now both Xorg and the wireless adapter work together:

ra0-20060113.png

Interestingly I now have even more things doubling up on the same IRQs, but it nonetheless works. I've not yet figured out what perturbs the signal strength, but it fluctuates wildly, at times dropping down as low as 66%.

Posted by Mark at 08:19 PM | TrackBack

January 12, 2006

New hardware breaks video device

No idea, and cannot find a clue. When I start Ubuntu 5.10 after having plugged in the wireless card, the wireless card is recognized and works fine (although not on 5.04). But my xorg.conf doesn't work. Something weird is happening with my video. From dmesg:

[4294758.173000] agpgart: Found an AGP 2.0 compliant device at 0000:00:00.0.
[4294758.173000] agpgart: Putting AGP V2 device at 0000:00:00.0 into 4x mode
[4294758.173000] agpgart: Putting AGP V2 device at 0000:02:00.0 into 4x mode
[4294764.798000] NVRM: RmInitAdapter failed! (0x12:0x3a:1167)
[4294764.798000] NVRM: rm_init_adapter(0) failed

Xorg fails with a message that tells me it cannot read or write or something from /dev/nvidia0. Of course the permissions on that device are crw-rw-rw-, so something weird is going on. I reinstalled the packages thinking perhaps something had gotten corrupted, but that didn't help.

Too tired to think about it now. Going to bed.

Posted by Mark at 11:17 PM | TrackBack

January 11, 2006

Trackback and comment spam down

If you have a blog, you no doubt get spam in the comments and through trackbacks. By the way, the difference between comments and trackbacks are that comments involve somebody leaving a note on one of your blog entries, whereas trackbacks are links back from someone else's site on which they cited your entry. MovableType 3.2 has a ranking system for filtering out the comment and trackback spam. It works fairly well.

A while ago, however, I was submerged with trapped spam, hundreds of comments and trackbacks each day. That seems like a lot for a blog only a handful of people read. I felt compelled to get rid of junk comments and trackbacks by hand, for fear of slighting someone out there who left a legitimate comment that got filtered.

For some reason the comment spam recently dried up. Even the trackback spam is at a trickle of a few per day. Why is that?

Posted by Mark at 09:17 PM | Comments (1) | TrackBack

January 08, 2006

Voice over IP

The number has been attributed for the Freebox. I picked up the phone and got a dial tone.

Dad had called this morning already. I called Mom's. Dana answered to say that Mom is out today on a visit. We talked a while and had a fully acceptable connection, with only a few strange noises. Seems to be usable.

Posted by Mark at 04:47 PM | TrackBack

January 06, 2006

Online robbery

We're switching our DSL provider to Free.fr. It's great.

I don't actually have the hardware. Online I see the box is theoretically somewhere in the same department as Lyon according to the French postal tracking site. So there's no way using what they've done for me, yet.

They have however managed to bill me already. That must've been a part of the fine print I misunderstood. The part where in a regular human language it would be obvious that they don't actually have to provide you a service to start taking money out of your bank account.

Posted by Mark at 08:52 PM | TrackBack

January 02, 2006

Spring cleaning needed

Looked in my home directory on the PC, since I'm still thinking of backing it up and moving stuff.

$ du -hs .
20G .

Only 10G of that is SVCDs, and 6.7G is ripped music. But that still leaves more than 3G. How did I accumulate all that stuff?

Posted by Mark at 11:22 PM | TrackBack

December 27, 2005

DSL envy in Desvres

We arrived, mostly without a hitch, just half an hour late in Lille.

Colette was wondering about DSL at home. Cegetel wanted to give her a 2 Mbit line. I looked up their stats at DegroupeTest.com. Only 458 m by wire from the commuter, 6 dB attenuation! They could have at least 8 Mbits, and here we are on dial-up.

Unfortunately the line cannot be degrouped yet. Still I should move up here. My connection would be almost as good as it is in the office.

Posted by Mark at 07:23 PM | TrackBack

December 24, 2005

/home at home

When I bought a new disk for the PC at home, I partitioned it right away and put the new Ubuntu on half of the new disk, leaving the old Ubuntu on the other disk. Then instead of using the existing /home, I started a new one on the new partition, leaving myself 40 GB for "files" on a second partition of the new disk.

It later occurred to me that I've been taking the wrong approach for years. $HOME never really changes. You carry it with you. Systems come and go every six months or so. Ideally I'd try out a new system probably every three months, maybe more often. The ones I want to try out on this PC, but haven't because I made such a hash of partitioning, include OpenSolaris -- Have we incorporated enough of the GNU stack yet? -- Gentoo, FreeBSD, a recent Fedora, maybe something exotic. Yet our $HOME files need to be in no danger just because I want to try out a new OS.

Backup would be simplified radically as well. I used to back up software that came from someone else. Only recently have I realized that unless you're the one making the software, you should not bother backing it up. Save your software. Save your config. Save your content. The rest you can download from somewhere. If you cannot download the software from somewhere and yet you rely on it, are you sure you're doing the right thing?

Posted by Mark at 08:34 AM | TrackBack

December 23, 2005

Still no Gmail, but only under Ubuntu 5.04

It's strange, I can no longer get to Gmail under Ubuntu 5.04. What is that? Some strange component problem under Firefox 1.0.7?

The interesting thing is, my Inbox gets read when I go to the personalized home page at Google. But when I try to get through to Gmail, I cannot do it.

Posted by Mark at 09:39 PM | TrackBack

How to of the day

Went to the top page of Google, where you can get a personalized home page, kind of like the portal we have at work. One of the categories I'd added was "How To" of the Day.

Today my top how to of the day is How to Write a Resignation Letter. They must've noticed we didn't get a half day off for Christmas or New Year's this year.

Posted by Mark at 04:57 PM | TrackBack

December 19, 2005

Browser as resource hog

You use it all the time, and it burns through lots of resources. Your web browser.

As I write this entry, I have one browser window open. I'm also downloading a file. Alongside that, I'm converting some digital video files to MPEG.

Firefox is burning through about 1/3 as much CPU as ffmpeg! While I use it to enter some text in a form, and download a single MP3 file. In the last half hour or so, Firefox has burned through about 8 minutes of CPU time, 4x what Xorg uses.

Posted by Mark at 09:30 PM | TrackBack

December 18, 2005

Video capture, part XXIII

Transcode, by the way, is hanging converting raw digital video instead of .avi files. Not for each file, but since the only solution I've found that works is kill -9 pid, I need to perhaps export these clips to .avi. Too bad I cannot fix it before Mom goes home.

Posted by Mark at 05:25 PM | TrackBack

Video capture, part XXII

Many CLI video processing tools are worse than ldapsearch. Check out the list of options in the man page synopsis for transcode:

transcode(1)                                                      transcode(1)

NAME
transcode - LINUX video stream processing tool

SYNOPSIS
transcode [ -i name ] [ -H n ] [ -p file ]
[ -x vmod[,amod] ] [ -a a[,v] ] [ --dvd_access_delay N ]
[ -e r[,b[,c]] ] [ -E r[,b[,c]] ] [ -n 0xnn ] [ -N 0xnn ]
[ -b b[,v[,q[,m]]] ] [ --no_audio_adjust ]
[ --no_bitreservoir ] [ --lame_preset name[,fast] ]
[ -g wxh ] [ --import_asr C ] [ --export_asr C ]
[ --export_par N,D ] [ --keep_asr ] [ -f rate[,frc] ]
[ --export_fps f[,c] ] [ --export_frc F ] [ --hard_fps ]
[ -o file ] [ -m file ] [ -y vmod[,amod] ] [ -F codec ]
[ --avi_limit N ] [ --avi_comments F ] [ -d ]
[ -s g[,c[,f[,r]]] ] [ -u m[,n] ] [ -A ] [ -V ] [ --uyvy ]
[ --use_rgb ] [ -J f1[,f2[,...]] ] [ -P flag ] [ -D num ]
[ --av_fine_ms t ] [ -M mode ] [ -O ] [ -r n[,m] ]
[ -B n[,m[,M]] ] [ -X n[,m[,M]] ] [ -Z wxh[,fast] ]
[ --zoom_filter str ] [ -C mode ] [ --antialias_para w,b ]
[ -I mode ] [ -K ] [ -G val ] [ -z ] [ -l ] [ -k ]
[ -j t[,l[,b[,r]]] ] [ -Y t[,l[,b[,r]]] ]
[ --pre_clip t[,l[,b[,r]]] ] [ --post_clip t[,l[,b[,r]]] ]
[ -w b[,k[,c]] ] [ --video_max_bitrate ]
[ -R n[,f1[,f2]] ] [ -Q n[,m] ] [ --divx_quant min,max ]
[ --divx_rc p,rp,rr ] [ --divx_vbv_prof N ]
[ --divx_vbv br,sz,oc ] [ -c f1-f2[,f3-f4] ] [ -t n,base ]
[ --dir_mode base ] [ --frame_interval N ] [ -U base ]
[ -T t[,c[-d][,a]] ] [ -W n,m[,file] ]
[ --cluster_percentage use ] [ --cluster_chunks a-b ]
[ -S unit[,s1-s2] ] [ -L n ] [ --import_v4l n[,id] ]
[ --pulldown ] [ --encode_fields ] [ --nav_seek file ]
[ --psu_mode ] [ --psu_chunks a-b ] [ --no_split ]
[ --ts_pid 0xnn ] [ --a52_drc_off ] [ --a52_demux ]
[ --a52_dolby_off ] [ --print_status N[,r] ]
[ --progress_off ] [ --color N ] [ --write_pid file ]
[ --nice N ] [ --accel type ] [ --socket file ]
[ --dv_yuy2_mode ] [ --config_dir dir ] [ --ext vid,aud ]
[ --export_prof S ] [ -q level ] [ -h ] [ -v ]

Makes you want to do something else, doesn't it? Like lie down and forget the whole thing.

Posted by Mark at 03:52 PM | Comments (2) | TrackBack

December 15, 2005

Gmail timing out

Not sure why, but I'm having regular trouble getting through to Gmail from home the last couple of weeks. Firefox waits and waits, then times out. Ping's okay:

$ ping gmail.google.com
PING gmail.l.google.com (72.14.205.107) 56(84) bytes of data.
64 bytes from 72.14.205.107: icmp_seq=1 ttl=239 time=167 ms
64 bytes from 72.14.205.107: icmp_seq=2 ttl=239 time=167 ms

--- gmail.l.google.com ping statistics ---
2 packets transmitted, 2 received, 0% packet loss, time 1000ms
rtt min/avg/max/mdev = 167.005/167.055/167.105/0.050 ms
$

Maybe I should Google for the answer? (Update: "gmail not responding" first hit is for a mail Subject: slapd not responding to sasl_bind. Hmm.)

Posted by Mark at 10:28 PM | TrackBack

December 12, 2005

Quick slideshow, part II

Quick slideshow, but not too quick. I decided just to guess at the number of frames to hold each photo. Five, about 5 photos/sec, was too quick. Half that fast is okay for us. We have lots of duplicates. Laziness gave me this:

#!/bin/bash
for item in `ls ${1}/dsc[0-9]*.jpg`
do
convert $item /tmp/`basename ${item} .jpg`.ppm
done

rm /tmp/list.photos
touch /tmp/list.photos
for j in `ls /tmp/*.ppm`
do
for i in 1 2 3 4 5 6 7 8 9 10
do
echo $j >> /tmp/list.photos
done
done

cat /tmp/list.photos | xargs -n1 cat | ppmtoy4m | mpeg2enc -o $2.m1v

I wonder if it's worth rewriting for configurability. Takes longer to run the script than to write it. The first MPEG compilation is diane.m1v (16 MB).

Posted by Mark at 09:42 PM | TrackBack

Over to bash at work

Switched my default shell to bash last week. For quite a while I've been using tcsh at work, bash at home. Wonder why I was doing that. Must've been something about what was installed by default.

Posted by Mark at 10:51 AM | TrackBack

December 11, 2005

Quick slideshow?

For each of the children, we have a bunch of JPEG photos all the same size, taken with the camcorder.

I found some doc on how to convert JPEG to MPEG, and wrote a short conversion script. (All the photos start with "dsc")

#!/bin/bash
for item in `ls ${1}/dsc*.jpg`
do
  convert $item /tmp/`basename ${item} .jpg`.ppm
done

ls /tmp/*.ppm | xargs -n1 cat | ppmtoy4m | mpeg2enc -o mpegfile.m1v

Trouble is, this is one image per frame of the MPEG. At that rate it's just a mess. Is there a trick to slowing it down? Just cat each image a few times?

Posted by Mark at 05:36 PM | TrackBack

December 08, 2005

DSL eligibility

Tested our telephone lines at DegroupTest.com. Their tests show the same results as what I've seen in the past, which is that France Telecom can give us nothing better than 512Kb. 58.76 dB attenuation on both lines, each 5202 meters from the hookup.

The sys admins at work are still suggesting Free.fr instead of France Telecom, mainly because Free's pushed the DSL envelope lots harder than France Telecom. Pierre says Free's starting to introduce 100Mb lines in metropolitan areas. The pricing seems to be identical. Maybe I'll go ahead.

Posted by Mark at 08:18 PM | TrackBack

December 02, 2005

Public beta, part II

Gmail's unavailable, and has been for the past hour. Cannot get to Google.com either. Not sure what the deal is, nor am I going to waste my time trying to figure it out.

The other glitch right now is with the printer under Ubuntu. From 5.04 it prints images from Firefox, no problem. From 5.10, it doesn't. The stuff in syslog looks like this:

Dec 2 21:10:16 localhost kernel: [4295502.794000] drivers/usb/class/usblp.c: usblp0: nonzero read/write bulk status received: -110 Dec 2 21:10:16 localhost kernel: [4295502.797000] drivers/usb/class/usblp.c: usblp0: error -110 reading printer status

Something Google.fr took me to at LinuxQuestions.org suggests I should try /etc/init.d/hotplug restart.

Posted by Mark at 09:32 PM | TrackBack

November 29, 2005

How to lower unemployment

As everyone knows, but nobody says out loud, the primary mission of the French ANPE is to reduce the numbers of people registered as unemployed... by whatever means necessary. Basically the most difficult way of getting you out of the stats is finding you a job, so they don't try to do that. It's considerably easier to eliminate you by having you jump through silly bureaucratic hoops until you give up or get eliminated for getting caught up in one of the hoops.

Nathalie worked 7 hours this month. That's not yet enough for her to be completely off the roll. So the first thing they tried, since they know from the information they required earlier that she's a mother of three children who are at home all Wednesday, was to set up a mandatory meeting in the middle of a Wednesday. They claim it's impossible to change the meeting time once it's arranged. By definition you are not employed, so therefore your schedule is wide open. They don't even have to ask you before requiring your presence at a certain time.

Another way they get you is to make it technically difficult to get the papers you must have in order to prove that you jumped through the useless hoops. Assedic.fr has some unpleasant features, such as a useless pop-up on the cover page that Firefox of course blocks by default, and scrolling text that gets garbled in Firefox. If you're out of work on the dole but want to be able to use the site, you still have to pay the Microsoft tax.

But they outdo themselves on the page that handles getting a version of the monthly declaration you need to avoid getting thrown off the list. You do your declaration online, then they generate a PDF for your printing. Unfortunately neither Adobe Acrobat nor xpdf can actually read or print the content of the generated file. And oh, by the way, you can only go through the process and generate the file once.

When you contact their tech support, they tell you how to configure your browser to accept the pop-up.

Posted by Mark at 08:42 PM | Comments (1) | TrackBack

In search of

Another good article today, this time in the Column Two blog, about intranet search. Take the "few practical steps" from this one, cross it with my last post, and see what you get.

Posted by Mark at 03:06 PM | TrackBack

Public beta

WSJ.com has an article with one of the best quotes I've seen in a while, concerning public beta software:

"I deplore it as a consumer; I admire it as a marketing professional," said Peter Sealey, a marketing professor at the University of California at Berkeley and former chief marketing officer at Coca-Cola Co. "I can't come up with anything else in the entire marketing world where marketers knowingly introduce a flawed or inadequate product [and] it helps grow your user base."

Somehow professor Sealey manages to explain the fundamental design problem of our civilization in only two sentences. It took Philip K. Dick a whole book.

Posted by Mark at 02:56 PM | TrackBack

November 14, 2005

See what isn't there

Norm Walsh mentioned this curious optical illusion in his blog. Try it.

Posted by Mark at 08:50 PM | TrackBack

Web 2.0 now official

Dad sent a link to an Investor's Business Daily editorial in which the basic message is that Web 2.0, whatever that is, is now a reality because Bill Gates wrote email saying it's coming.

So it's now official. You can go on paying 86.49 times earnings for Google, because This Time It's Different!™

Posted by Mark at 02:42 PM | TrackBack

November 09, 2005

Schema Repository, part XV

It wasn't until I turned off the computer and quit looking at the code that I could see my approach was the wrong one in many spots. Nathalie asked me what I was doing.

"I'm thinking."

"I can see that. What are you thinking about?"

"Storing LDAP schema and documentation in a repository."

"What?"

Until now, I have had separate attributes to store the pieces of the object definitions. In other words, I have attribute types like schema-oid, schema-name, schema-must, schema-may. But I also store the raw definition. Not sure why I started down that road. It seems like I could have parsers for raw definitions that give the same sort of API as the attribute types would, but without the hassle of storing duplicates.

Then I have one big blob called schema-fulldesc for most of the documentation. And I think my schema-example attribute type is single-valued. Finally, my collections of objects don't have a memberOf functionality. I only saw I missed that when I started looking at man pages, wondering, "Okay, which collection would hold this one?" I don't think the attribute type entries even reference the container object class entries.

Oh, well. The first time you go through it is an exercise in figuring out more explicitly what you're trying to do. In that respect, writing software is much like any other form of writing. I need to block out time to revise.

Posted by Mark at 06:25 PM | TrackBack

November 04, 2005

GUIs for spammers

A couple of my accounts got spam today from somebody trying to spoof eBay.com, saying that the account is locked, I cannot use eBay until I click the link, etc.

The link they wanted me to connect to is port 680 on the host with IP address 218.236.22.164.

$ host 218.236.22.164
Host 164.22.236.218.in-addr.arpa not found: 3(NXDOMAIN)
$ nmap 218.236.22.164

Starting nmap 3.81 ( http://www.insecure.org/nmap/ ) at 2005-11-04 21:45 CET
Interesting ports on 218.236.22.164:
(The 1650 ports scanned but not shown below are in state: closed)
PORT STATE SERVICE
135/tcp filtered msrpc
139/tcp open netbios-ssn
445/tcp filtered microsoft-ds
680/tcp open unknown
780/tcp open wpgs
1026/tcp open LSA-or-nterm
1027/tcp open IIS
1029/tcp open ms-lsa
1433/tcp open ms-sql-s
3128/tcp filtered squid-http
3389/tcp open ms-term-serv
4444/tcp filtered krb524
17300/tcp filtered kuang2

Nmap finished: 1 IP address (1 host up) scanned in 101.956 seconds
$

So other than them seemingly running Windows, what's up? I found a page that says a virus called RTB 666 uses that port. Looks like these folks can set up trojans running on other people's machines, then get the account info back through those trojans.

RTB 666 seems to have a GUI client in Polish. Here's the half-size thumbnail:

screen capture

Click the image to see the MegaSecurity.org screenshots.

Posted by Mark at 09:53 PM | Comments (1) | TrackBack

October 30, 2005

StyleCatcher nevermind, part III

Jason Lefkowitz's site and mine got upgraded at the same time, because my brother's at the same hosting service. Jason kindly left me a note to his entry:

Turns out that StyleCatcher has problems with old versions of the Perl module libwww-perl (aka LWP). My box was running LWP version 5.64, which is an old version, vintage 2002 — the latest one is 5.803, which is what 6A had been using for testing.

Huh?

$ head movabletype/extlib/LWP.pm
#
# $Id: LWP.pm,v 1.118 2002/02/09 18:45:42 gisle Exp $

package LWP;

$VERSION = "5.64";
sub Version { $VERSION; }

require 5.004;
require LWP::UserAgent;
# this should load everything you need
$

So if they're testing with 5.803, how did I get 5.64 with the upgrade I downloaded from their site?

Library versioning makes all software people look brilliant. It's the software engineering equivalent getting your picture taken while picking your nose.

At least StyleCatcher now sorta works.

UPDATE: Beckett is great. I feel like Wonder Woman with all these stars.

Posted by Mark at 09:31 PM | TrackBack

October 25, 2005

fortune

When I first tried Slackware, one of the things I liked was the message of the day being a fortune cookie.

You can often get that today from the command line. For example:

$ fortune
If we do not change our direction we are likely to end up where we are headed.

It's somehow reassuring that the sys admins didn't forget to install fortune on the SunRay. Prevents you from taking your situation too seriously. I wonder if they avoided installing the "offensive" cookies.

For a twist on this, you can always ask the oracle your question of the day.

Too bad they don't give you the fortune cookie right there when you login.

Posted by Mark at 02:18 PM | TrackBack

October 24, 2005

Vpnc works fine

If you don't want to have to recompile the kernel to use the Cisco VPN 3000, vpnc seems to work. Two things to keep in mind when using it:

  1. You must be root (or sudo) when playing with your network interfaces, and so you must be when using vpnc.
  2. You must open port 500 to isakmp UDP packets when you connect to the VPN.

I Googled for "cisco vpn itables" and found what I'd forgotten since I read Cisco's install doc for the VPN 3000 client.

Let me know if you want my conf or script to use vpnc.

Posted by Mark at 09:19 PM | TrackBack

October 23, 2005

Gcompris

Since wine's hit-or-miss, I've installed something called Gcompris. Tim's going to give it a try. We'll see.

Posted by Mark at 10:04 AM | TrackBack

Bad wine, part III

A CD for 2/3 year-olds works under wine. It's called, appropriately enough, Maternelle : Petite section, with Lisette the duckling as the main character. Emma tried that for a while, but it's too easy for her.

Posted by Mark at 09:37 AM | TrackBack

Bad wine, part II

Tibili won't work under wine, apparently. The only debug errors I see are about missing fonts.

The main symptom comes when I start Tibili.exe under wine. It asks to load a Director movie file. I found one on the CD that's an ad for Wanadoo, the France Telecom ISP. So it runs that and quits shortly after.

Installation proceeded without a hitch. Just doesn't work, that's all.

The same is true for a promotional CD from Danone, who spent a lot of money to sell liquid yoghurt at extortionist prices under the Actimel brand. I'm surprised they didn't get censored for claiming roughly that their product is a miracle cure that let's you eat bad meals and nevertheless stay healthy.

Anyway, wine's missing emulation of a DLL:

err:module:import_dll Library MFC42.DLL (which is needed by L"C:\\Program Files\\Equipe Actimel\\JOUER.exe") not found

Changing the Windows environment (W98 instead of the default) doesn't help. I'm sort of stumped. Going to have to tell Emma that I give up.

Posted by Mark at 08:39 AM | TrackBack

October 21, 2005

MovableType & junk removal

MovableType 3.2 does a pretty good job removing junk. It also holds some comments or trackbacks until the author approves. We have a lot less raunchy porn advertisements to read that way, and it robs the guys doing that from improving their ratings at Google at our expense.

But here I am, responding to a comment on my own blog, and:

Thank you for commenting. Your comment has been received and held for approval by the blog owner.

In the same browser the password manager lets me login without asking.

Posted by Mark at 09:24 PM | TrackBack

Bad wine

A long time ago, I built wine, the WINdows Emulator, for Emma so she could run Tibili, which is a program for children learning to read French. Back then I got it working. Not so this evening.

I tried creating symlinks to the wine drives, but to no avail. Every time I start Tibili.exe, xorg dies and gdm restarts it. A quick way to kill your X session. This is the version of wine I get with Ubuntu 5.04.

Posted by Mark at 09:06 PM | TrackBack

October 20, 2005

LDAP client SDKs

Ever since I found out one day that the Directory SDK for Java docs were some of the hardest hit straight HTML on docs.sun.com, I guess I've wondered which Java-based toolkit has the most followers. But how would you really know?

Googling for ldap java sdk brings up the Netscape-branded doc to the Directory SDK for Java, with the 4.0 Programmer's Guide at docs.sun.com close behind. (Both docs cover essentially the same code base.)

If I change that to directory java sdk, the Java standard page pointing to JNDI rises further up the stack of results, but the Directory SDK for Java remains in the "I'm Feel Lucky" spot.

A Googlefight of JNDI vs. Directory SDK for Java leaves JNDI on the ropes, but only by about 10%, though JNDI kicks LDAP Java SDK's butt.

Back at Google, in their directory, Directory SDK for Java is the highest-ranked page for all of Computers > Programming > Languages > Java > Development Tools.

People are going to be making jokes about LDAP client code like they used to joke about Cobol before the year 2000. Incidentally, Java demolishes Cobol in a Googlefight by a ratio of about 346 to 5.

What to conclude? There are still lots of developers out there at least checking out, and linking to the doc for the Directory SDK for Java, which is still version 4.x. (Version 4.x started in 1998 and has been reasonably stable.) Funny that would last so long, even with JNDI being handled through the JSR process and being a part of the Standard Edition SDK, so available everywhere, for the last couple of years.

As you may or may not know, JNDI is a more generic API that lets you do directory client applications, but also applications that work with other sorts of naming systems, like DNS or file systems. Directory SDK for Java sticks close to the LDAP protocol, so developers who want to do just directory clients may find it a closer fit.

Posted by Mark at 08:30 PM | TrackBack

October 19, 2005

Should we have frequency monopolies?

Forbes.com is running an article about Eben Moglen's challenge that the FCC broadcast frequency allocation model is outmoded:

By using open-source software and low-powered “mesh networks” that can sniff out open frequencies and transmit over them, Moglen says, “we can produce bandwidth in a very collaborative way,” including transmitting video and telephone conversations that would normally ride on commercial networks.

Okay, as long as those frequencies are truly open. It may be that new software, by using bandwidth differently, can work around the problem, sort of like you can have DSL and voice telephone on the same line. In any case, it would be nice to see broadcast bandwidth financed by big advertisers come under competition from smaller transmitters and peer-to-peer communications.

Posted by Mark at 08:27 PM | TrackBack

October 15, 2005

Ubuntu, part XIII

Downloaded the new Ubuntu, which I intend to install soon.

This one is called Breezy Badger. Maybe it's a joke about Linux on a Dead Badger, I don't know. But I do find it reassuring to get a distro from people as oblivious to marketing as the folks at Debian, where Ubuntu comes from. Let them waste their time focusing on what they're doing rather than on packaging to attract people who pay attention to that sort of thing.

Maybe we should get a little less slick at work. Our site's looking too cool these days.

Posted by Mark at 03:22 PM | TrackBack

September 30, 2005

Xorg hangs..., part II

... and it could be Mozilla. At least it happened while I was running Firefox. Then I got this in dmesg:

kernel BUG at mm/rmap.c:483!

There's a sort of stack trace after that. What does one do with that?

Posted by Mark at 09:14 PM | Comments (1) | TrackBack

September 27, 2005

Ubuntu, part XI

After some fiddling with xorgconfig, I got Xorg functioning on the laptop from work. For some reason the Debian X configuration program didn't associate the Trident video card with the trident driver by itself, perhaps because the particular card in the laptop is something called XP4m32, which isn't in the database. As Gilles said of laptops, the hardware's just too specific.

The internal wireless hardware is recognized out of the box, but I have no wireless network at home, so haven't tried to configure it.

Whoever said "Go" for that annoying little blue mouse nub in the middle of the keyboard needs to review the test plan. Ostensibly that thing is so you can use the mouse even if you cannot figure out how to use a touch pad. (In other words, it's useless even from a design point of view.) In fact it's there to send the mouse pointer in random directions firmly to the edges of the screen so you get to use X without a mouse.

Under Windows Ludo showed me where to look to deactivate the blue nub. Need to Google for the answer in Xorg terms.

Posted by Mark at 06:48 AM | TrackBack

September 26, 2005

Xorg hangs...

...and it seems to be related to something in Mozilla. That bug I found still seems to be stuck in assigned. Red Hat was telling people to upgrade.

Xorg hangs fairly regularly when we're using Firefox or Thunderbird. I get a situation where Xorg is eating up 99.x% of CPU time according to top. Usually I can plug a laptop in and dhcpd still responds, so the laptop gets a connection and I go in over ssh to clean up.

Cleaning up means:

  1. Find the pid for xorg and kill -9 it.
    At this point everything in the session gets lost but I still...
  2. pkill -u <user> to finish off the processes still alive after xorg dies.
  3. If gdm doesn't take give me another login screen, kill any extra gdms, and sudo /etc/init.d/gdm restart.
    Regularly when xorg dies from an abrupt kill, Gnome display manager gets confused.

The painful parts of this are not only that you lose whatever you were doing, but you also have to get a laptop out and boot it just to do the maintenance.

Mr. Bones suggests commenting out Option "RenderAccel" "true" in xorg.conf. I'm going to try that.

Posted by Mark at 09:54 PM | TrackBack

September 25, 2005

Dazed and confused under suexec, part II

Found some old mail from my brother to a FAQ at the place hosting this site. It of course explains the whole thing with suexec and having permissions like:

$ find . -print -exec chmod go-w {} \;

It would be nice then if my umask were not 0002. That means any time I unpack a tarball all the CGI file permissions are going to be wrong by default. The only one there sharing my group is my brother anyway. I wonder why I have to be so worried about that. Is it easier to crack our group than to crack our passwords?

Posted by Mark at 06:02 PM | TrackBack

September 22, 2005

BIOS problem, part II

Hmm. That was 7 times in a row. After the first 4, I quit doing resets and turned everything off. The last three times Xorg seized up almost immediately after the start of the session.

Now I've turned the printer power off. So far that seems to have worked. Strange.

Posted by Mark at 08:12 PM | TrackBack

BIOS problem

Something's not right since I plugged the printer in. Every few boots, the PC goes batty before I even get into the system, when the BIOS is still taking inventory. The screen gets all messy and everything hangs, even before grub starts.

I wonder if there's something funny about the USB setup I have. Not sure how to debug it.

Posted by Mark at 08:04 PM | TrackBack

September 18, 2005

StyleCatcher nevermind

Seems like I've run into whatever Jason Lefkowitz encountered. I've gone back to all defaults in an attempt be download the styles, but nothing doing.

Shucks, I guess I'll have to figure out manually how to download a new, gaudy style each day.

Posted by Mark at 05:15 PM | Comments (1) | TrackBack

MovableType upgrade

Just upgraded to 3.2. Must've been the permissions thing last time. Seems to have gone okay.

Posted by Mark at 03:32 PM | TrackBack

Software glitch weekend?

Firefox and Xorg did their thing and hung everything a few seconds ago. I couldn't even get and address from the DHCP daemon on the system to be able to go in there and kill Xorg.

Maybe I shouldn't be playing with any software this weekend.

Posted by Mark at 02:52 PM | TrackBack

Dazed and confused under suexec

Okay, I'm back. You may not have noticed, but I tried to update to MovableType 3.2.

Everything just broke. It appears there was a problem with who ran what. At least that's what's fixed it. I finally found warnings about all that in the server suexec doc. Everything was set up as my user, and my brother's user, which I guess is our real sys admin, is the identity the server's taking.

Couldn't the server just tell me that in the logs? This message is not exactly edifying to the vulgate:

Premature end of script headers: /path/to/test.cgi

This can mean, AFAIK:

What I really do not understand is that I had the same ownership and permissions as I have on some CGIs I wrote and the other ones work!

Security through obfuscation? What happens when a smarter fool comes along?

Posted by Mark at 02:23 PM | TrackBack

September 13, 2005

Ooops

Sometimes MT seems hung, but I guess it was just the trackback ping that got generated for the last entry. Never really understood those. Maybe I should turn them off.

Posted by Mark at 10:50 PM

Ubuntu, part X

Norm Walsh tried Ubuntu, and it seems to work for him.

He got the VPN up using the open source client. I didn't even realize such a thing existed, and was using the closed source tarball with the config files from our IT folks.

When I set up the open source client, I found traffic getting stuck at port 500. Know I've seen that before...

Posted by Mark at 10:50 PM

September 11, 2005

Ubuntu, part IX

There is one thing I ought to complain about: Firefox hangs.

Every once in a while, Firefox hangs, Xorg goes into neverneverland and eats up all the CPU cycles doing nothing. I can ssh in from the laptop and kill Xorg to have everything recover, but the only thing I can do on the PC itself is reboot. The keyboard just stops responding.

I cannot figure out why this happens, but it does. Ugly bug.

Posted by Mark at 09:57 PM

Ubuntu, part VIII

Huh? Things have changed since 2.4.21. I thought I couldn't get my CD to burn under Ubuntu last time. sudo cdrecord -scanbus shows only my scanner.

This time I'm less sleepy. Googling to an Ubuntu forum answer by somebody with the handle "skoal," I learned that you don't necessarily need one of those odd SCSI device numbers. You don't need hdc=ide-scsi at boot time. All you need is the regular /dev/hdc style device name and away it goes:

$ cdrecord -v dev=/dev/hdc wizom.iso

So, uhh, how the heck did I just burn a CD as me? Not even sudo. That was very strange, but it worked.

Posted by Mark at 08:39 PM

September 09, 2005

ML-1610

The ML-1710 driver doesn't seem to be working for the ML-1610 printer. Nathalie got something to printer earlier today. Since then, we've just been getting garbage and blank pages.

It's quite frustrating. Guess I'll have to reset the root password to know what it is just so I can install my printer. Cannot figure out how to do it with sudo.

Right now I cannot even get a test page to print properly... reinstalled the same driver, power cycled the printer and it worked. If only I knew why.

Posted by Mark at 09:13 PM

September 07, 2005

Vpnclient working again

I'd suspected as much. Whatever it was in dnsmasq and maybe ipmasq, everything worked okay when I stopped those before initializing the VPN modules.

$ sudo /etc/init.d/dnsmasq stop
Password:
Stopping DNS forwarder and DHCP server: dnsmasq.
$ sudo /etc/init.d/ipmasq stop
Disabling IP Masquerading...done.
$ sudo /etc/init.d/vpnclient_init start
Starting /opt/cisco-vpnclient/bin/vpnclient: Done
$ vpnclient connect ...

After that I could get my work email and so forth. Great.

Posted by Mark at 06:50 PM

September 06, 2005

Ubuntu, part VII

After formatting the file system around bad blocks, I installed Ubuntu on the Vaio. Works fine with xfce instead of Gnome, and so do dnsmasq/ipmasq.

Proof is that I'm on DHCP to the PC from the Vaio here writing this entry. All I need to do now is snake a cable further away and I can waste time writing blog entries while Tim plays TuxRacer or Emma plays games at Barbie.com.

Posted by Mark at 10:04 PM

September 02, 2005

Skipping bad blocks

The old Vaio laptop I have seems to have some bad blocks on the disk. Fabio suggested I make a file system that skips bad blocks. For example, with a Slackware install CD Fabio had lying around:

# mke2fs -c -c -j -m 1 -v /dev/hda1

This presumes the disk is already partitioned but nothing's been installed. The command doesn't seem to work with Ubuntu, because there's no badblocks command when you start installation.

It takes a lot longer to make the file system than to install the OS. Everything was proceeding quickly until it got to the bad part of the disk.

Posted by Mark at 05:04 PM

August 27, 2005

To upgrade or not to upgrade

Apparently the 3.2 version of MovableType is out. I'm wondering whether to take the time to upgrade.

It's surprising how easily I accept updates when they're done for me, compared to how uncomfortable I feel when contemplating them.

Reflecting on this, I notice that as long as the configuration state I manage and my data aren't involved, I don't care. Upgrade away.

Whenever my data and the configuration state I have to manage is involved, there's a slighly perceptible allergic reaction, and a pretty strong desire not to go first.

The bar remains high for new features and bug fixes.

Posted by Mark at 09:34 AM

August 25, 2005

Waste

After some incomprehension, I finally decided to examine the README for the Quake III demo.

icon.jpg

It turned out, I had at minimum to run the game as root (or sudo on Ubuntu). Also, I was linking with a library for... duh!... the wrong video card. So once I straightened that out, I could get it to work:

$ ls -l
total 4216
drwxr-xr-x  2 root root    4096 2005-08-23 23:18 demoq3
drwxr-xr-x  5 root root    4096 2005-08-23 23:18 Help
-rw-r--r--  1 root root    2854 2005-08-23 23:20 icon.bmp
-rw-r--r--  1 root root    6852 2005-08-23 23:20 icon.xpm
lrwxrwxrwx  1 root root      43 2005-08-25 22:25 libGL.so -> /usr/X11R6/lib/nvidia/libGL.so.1.2.xlibmesa
-rwxr-xr-x  1 root root 1851824 2005-08-23 23:20 libMesaVoodooGL.so.3.2
lrwxrwxrwx  1 root root      18 2005-08-23 23:20 MOVEDlibGL.so -> libMesaVoodooGL.so
lrwxrwxrwx  1 root root      22 2005-08-23 23:20 MOVEDlibMesaVoodooGL.so -> libMesaVoodooGL.so.3.2
-rwxr-xr-x  1 root root  554088 2005-08-23 23:20 q3ded
-rwxr-xr-x  1 root root  915988 2005-08-23 23:20 q3demo
-rwxr-xr-x  1 root root  915988 2005-08-23 23:18 q3demo.x86
-rw-r--r--  1 root root   13510 2005-08-23 23:20 README
-rwxr-xr-x  1 root root   16204 2005-08-23 23:20 uninstall

Now that I've not played in a while, however, the game just seems like a waste.

Posted by Mark at 10:31 PM | Comments (2)

August 20, 2005

Ubuntu, part VI

I got mail from a guy doing Ubuntu, also named Mark. Mark's asking for my "help in formulating a national government strategy on Free Software for South Africa."

Sounds like a good idea, but I'm not sure how I'd be able to help. I checked out the taskforce Wiki nevertheless.

IMHO, they're in serious danger already. Half the projects up there have left-leaning titles like "Labour Department" and "Public Services and Administration" although one is "National Treasury." But the real problem is you can search the entire front page without finding the words "military," "prison," "police," or even "market."

You'd be better off reading the latest Slashdot article about how British Soldiers Get Germ-Fighting Undies. Let's keep the focus on the important things in life.

Posted by Mark at 11:32 AM

August 18, 2005

Rsyncing to backup this blog

Troy Johnson wrote some doc on Using Rsync and SSH to backup stuff over the Internet.

Thanks, Troy. Seems to work very well.

This also gets me thinking about the kind of documentation you really want to find when you cannot (or in my case are too lazy to) figure out for yourself how to get things to work.

We do read those man pages, but they're not a very gentle introduction to anything.

Posted by Mark at 09:38 PM

Ubuntu, part V

Looks like I spoke too soon about getting the VPN working.

In trying to prepare my second Ethernet card so I could use dnsmasq when plugging in laptops, I somehow overrode whatever was working before. I can still connect through VPN, but then none of the name services work. I cannot connect to anything.

Posted by Mark at 09:57 AM

August 17, 2005

JavaScripting a feed into static HTML

I'd been idly thinking about how to include a feed into an otherwise static page. Then I ran across RSS-xpress.

You add a smidgeon of script to the static page and your browser loads a table containing the RSS feed, such as your blog, from RSS-xpress. You can handle the formatting through CSS, and it's in a div if you want columns for example.

RSS-xpress is pretty limited. You configure only the formatting, and you get whatever's in the feed. No other control, since the browser's just getting the table generated from their server. But it sure doesn't require much development on your part.

Posted by Mark at 08:15 AM

Random playlist, part III

Looks like random is really not what I want.

Even if I were to add a line to normalize song volumes, some music just does not work during my commute at 25-30 kph. At that speed, which is not particularly fast, the wind noise drowns out most subtleties. Any music for which dynamics are crucial just dies. Maybe I should skip music while riding and listen to audiobooks.

Posted by Mark at 07:52 AM

August 16, 2005

Finding things in blogs

Shel Holtz wrote an article for webpronews.com on the difficulty of finding things in blogs. One of Shel's examples:

Sun Microsystems keeps all of its blogs on one page, but they're not organized in any way that makes sense. You can see recent posts to any blog along with a listing of "Hot Blogs" (dear God, when are we going to see an end to the overuse of the word "hot" on the web?). But if you're looking for a blog on, say, Java Studio Creator, you're back to using the search engine.

Dude, if you know what you're looking for, why are you browsing? Would you force yourself to read a whole book if you know you'll find the keyword you're looking for in the index?

You can get right to things at Google. Check out Java Studio Creator site:blogs.sun.com.

IMHO, the way to make information easier to find online is the present it in a standard format the bots can index, and maybe, if necessary, keep the page chunks from being too big for human beings to handle. Unless you're writing a web based application for specific user tasks, why do you presume someone could know beforehand what somebody else is going to be trying to find?

Posted by Mark at 08:36 AM

Random playlist, part II

Hmm. My playlist.py script was fun to write, and interesting as a novelty, but needs work. My collection is so Frank Zappa heavy for instance that in the first half hour of music this morning about 20 minutes were Zappa. Maybe I need a system of weighting somehow. But how?

Also, perhaps I should handle entire albums rather than songs. It's clear that most of the listening I've done since CDs come out has been albums in their entirety. Furthermore, if I gradually add audiobooks, then it makes less sense to split things up randomly by audio file.

Posted by Mark at 08:25 AM

August 15, 2005

Random playlist

I'm playing around with my MP3 reader and python. Python's a good language for handling lists and dictionaries and so forth, but I don't know it yet. So it took me some time to figure out where the pieces were for a short program that generates a random playlist of the size that fits onto my MP3 reader.

import glob
import os.path
import random
import shutil

MUSIC_SRC = "/home/mark/music/*/*/*.[mw][pm][3a]"
MAX_SIZE = 516538368L # Available blocks on my MP3 player when empty
MUSIC_DST = "/media/usbdisk/"

files = glob.glob(MUSIC_SRC)

list = {}
for filename in files:
list[filename] = os.path.getsize(filename)

playlist = []
totalsize = 0
maxsize = MAX_SIZE
while files:
song = random.choice(files)
files.remove(song)
totalsize += list[song]
if maxsize > totalsize:
playlist.append(song)

number = 1
for song in playlist:
copyname = MUSIC_DST + str(number).zfill(3) + " - " + os.path.basename(song)
shutil.copyfile(song, copyname)
number = number + 1

I've had one problem with this:

Traceback (most recent call last):
  File "playlist.py", line 29, in ?
    shutil.copyfile(song, copyname)
  File "/usr/lib/python2.4/shutil.py", line 48, in copyfile
    fdst = open(dst, 'wb')
IOError: invalid mode: wb

Huh? Forget it. I'm going to bed.

Posted by Mark at 10:58 PM

August 14, 2005

Too much

Just decided to unsubscribe from Google News as an RSS feed. Why would I do such a thing?

It's been almost 24 hours since I last looked at new blog entries on Bloglines.com. Google News had 200 new entries. Too much of a mediocre thing.

Posted by Mark at 09:01 PM

August 11, 2005

Finding things in email

At work we're being encouraged to clean up our mail. So I just got a copy of stuff that perhaps should be saved:

% du -hs .
 551M	.

I have another ~800M at home. How should I find content in this mine of precious data? (Thunderbird's okay, but the search interface is still frustrating harder to use than Google's.)

Most of it is work mail, hence I hesitate to upload it all to my Gmail account just to have it appropriately indexed. But that's the sort of search interface I want.

Posted by Mark at 09:05 AM

August 10, 2005

Failures

A couple of CDs simply refuse to be ripped.

The first one is Holst's Planets. Philips put a sort of plastic coating on the disc for the label that stuck to the carrying case and all over the place when the case was overheating in my car most probably. I suspect that is somehow the root of the trouble anyway.

The second one is King Crimson Discipline. The disc won't play in a CD player either. Curiously, it doesn't appear damaged in any way. Maybe heat melted something.

I guess I shouldn't store compact discs in my car for very long.

Posted by Mark at 10:10 PM

August 08, 2005

Off the grid

I've been ripping my CDs so I can listen to them on the MP3 player. Most of the data is found almost instantly in the web database. One of the CDs was off the grid: Stockhausen's Mantra, a work for two pianists.

album cover

It's understandable, since Mantra requires a lot more concentration to enjoy than easier albums such as Stimmung or even Aus den Sieben Tagen.

Posted by Mark at 10:22 PM

August 07, 2005

The cost of free

Somebody got a post up on Slashdot, where I almost never go anymore though I do scan the feed, linking to an article about schools in Indiana using Linspire, which is a commercial distro of Linux, GNU tools, OpenOffice, etc.

Huh? That's like Ubuntu, but costs a whole lot more. (And, I notice Edubuntu now, too.) What're they doing that for?

Support, perhaps:

Registered Users, Members, and Insiders can submit inquiries to Linspire Support personnel.

Issues are submitted via an online support form. You will receive an immediate acknowledgment of your inquiry. Our Knowledge system will usually select and send some potential solutions to speed the resolution of your inquiry. These resolve most common issues, and sometimes more advanced ones.

But rest assured we won't leave you with only a "canned" automatic system response! Your inquiry will also receive prompt attention by a Linspire support professional whose goal is to respond to and resolve issues within three business days."

(Source: http://support.linspire.com/support_policy.php)

I guess if you're the person making The Big Decision, you want the comfort of knowing some kid just out of college working at Linspire will get back to you in three days after Googling for the answer and trying out a few things, rather than you having to explain to your boss or the people who elected you or whomever that you went Googling for it yourself and spent 3 days trying to find the answer.

Of course, the Edubuntu people won't sell you support. They just put their contact information on that front page. None of that anonymous "one throat to choke" factor. Obviously taxpayer money is better spent on the commercial product.

Posted by Mark at 01:17 PM

August 05, 2005

Where Google's not enough

One of the blogs I scan sent me to an article on Recent Trends in Enterprise Search by Stephen Arnold. Arnold argues that:

Humans and human-like processes are needed to supplement or do certain types of taxonomy development, indexing, classification and analysis.

We do this at home all the time. Look in your kitchen cupboard. The dinner plates are probably stacked. The coffee cups are probably together somewhere else. The pots have their own place. The cutlery is similarly segregated. Where you actually put each class of item -- the specifics of your classification -- depends however on your particular way of handling meal preparation and cleanup.

You probably also organize your workspace somehow, and that organization corresponds to how you do your work, whatever it is. Your books and music are the same way. It's only when we want to share stuff that we need a shared organization protocol (Dewey decimal, Library of Congress, WWW, etc.), and indeed it's only when we want to share stuff that it makes sense to go through the trouble of using a protocol.

The more how we share information follows some sort of process, the more specific the protocol needs to be. If to use Arnold's example, "A lawyer working for the chief financial officer [needs] everything pertaining to a deal involving dozens of employees, several departments, and documents in all versions over a span of years," then a search engine using a PageRank algorithm won't be the right answer.

Yet I'm not sure humans need to do the indexing. Humans need to define the sharing protocol, though, and those doing the defining ought probably to be those very familiar with the process, able to define it formally or at least follow it intuitively in the same way as others follow it. That sharing protocol is likely to come out differently for each process in which information is shared.

So in the end I agree that Google's not always enough under our circumstances. Furthermore, maybe no finished general purpose search engine can be.

Trouble is, selling the idea that software has to be customized can be tough. As Linnea said, it used to be fast, cheap, or good, pick two. Now the choice is made: fast and cheap.

Posted by Mark at 07:41 AM

August 03, 2005

$5.5M?

We were talking about spam at lunch today. We haven't had so much at work lately. Then what follows comes in, with subject My Last Wish:

Good day,

My name is Mrs. Karen Grant I am a dying woman who have decided to donate what I have to you.
I am 59 years old and I was diagnosed for cancer about 2 years ago, immediately after the death of my husband, Who had left me everything he worked for.

I have decided to WILL/donate the sum of $5,500,000(five million five Hundred thousand dollars) to you for the purpose of charity work, and also to help the motherless and less privilege and also for the assistance of the widows .

At the moment I cannot take any telephone calls right now due to the Fact that my relatives are around me and my health status. I have adjusted my WILL and my Executor is aware I have changed my will; you and he will arrange for the change of ownership of the funds as it is presently deposited in a strong trunk box, and lodged. the box in a coded Security company whose name is withheld basically on security and confidential purposes and would only be released to (you).

I wish you all the best and may the good Lord bless you abundantly, and Please use the funds well and always extend the good work to others. Contact my Executor Michael Bacchus with this specified email m_bacchus@atmail.com with your full names contact telephone/fax number and your full address and tell him that I have WILLED ($5,500,000.00) to you and I have also notified him that

I am WILLING that amount to you for a specific and good work. I know I don't know you but I have been directed to do this. Thanks and God bless.

NB: I will appreciate your utmost confidentiality in this matter until the task is accomplished as I don't want anything that will Jeopardize my last wish. And Also I will be contacting with you by email as I Dont want my relation or anybody to know because they are always around me.

Regards,
Karen Grant

Out of idle curiosity, I check the mail headers. How about this?

X-AntiAbuse: This header was added to track abuse,
 please include it with any abuse report
X-AntiAbuse: Primary Hostname - blazecast3.rdns.net
X-AntiAbuse: Originator/Caller UID/GID - [32530 32531] / [47 12]
X-AntiAbuse: Sender Address Domain - virgin.net

Is that for some sort of mail client plug-in?

Posted by Mark at 06:52 PM

July 31, 2005

Can mount the MP3 player, part II

I've started ripping CDs to MP3s on my home system, which is still Red Hat 9. grip seems to work acceptably, although with a couple of gotchas that are no doubt fixed in a more recent version. I had to change a few things from the default (.ogg) encoding for things to work with the MP3 player:

grip-20050731.png

Notice that the encoder is lame, that I changed the -h to -p in the command line options, and move the spot where the encoded files go.

Unfortunately, I couldn't use %t in the output file names, so they're out of order compared to the disk and need to be renamed after ripping. Not sure what's going wrong there.

Posted by Mark at 10:03 AM

July 30, 2005

Can mount the MP3 player

Whereas I couldn't get my camera mounted, all I had to do for the MP3 player was plug it in, choose a mount point, and mount things:

$ ls /proc/scsi/
atp870u        ide-scsi       scsi           sg             usb-storage-0
$ sudo mount -t vfat /dev/sda1 /mnt/mp3player
$ ls /mnt/mp3player/
Buddy Guy              Johnny Cash - Unchained  settings.dat
Jimi Hendrix           nugscast_20050514.mp3    The Jimi Hendrix Experience
Johann Sebastian Bach  Pink Floyd               WMPInfo.xml

If that works so simply, then what the heck is going on with the HandyCam?

Posted by Mark at 05:45 PM | Comments (2)

July 28, 2005

Cannot mount the HandyCam

Cannot figure this one out, although I've been working at it for 3 hours, Googling included.

My system at home, kernel 2.4.21, recognizes the HandyCam when I plug it in over USB. But it doesn't associate the usb-storage driver with it. Nothing shows up in /proc/scsi except the stuff already there. So when I go to mount /proc/sda1 or any of those devices:

# mount -v -t vfat /dev/sda1 /mnt/camera/
mount: /dev/sda1 is not a valid block device

It's one of those occasions where I wish I knew how some of this actually works, because it must be really simple if you do.

Posted by Mark at 11:52 PM

July 21, 2005

Broken camera, part IV

Just when you think you know what you're doing, you realize you have no idea how this stuff works:

test-20057021.JPG

Proof that Tim's camera seems to have worked now that format pronounced it unusable. Unfortunately I'm confident I'll never recover those pictures from the top of the Sears Tower.

Posted by Mark at 01:18 AM

July 06, 2005

Mom's from space

I'm late to the party, having not looked at Google Maps since the announcement of its existence appeared. But you can get satellite views of the US.

Here's a link to Mom's neighborhood as seen from space. The yard does look pretty shady now.

Posted by Mark at 05:20 AM

July 05, 2005

Ad feeds

C|Net's got an article from the NYT about how advertising is now going into RSS. I've already seen ads from Sun coming through the feed from Slashdot.

There's no escape. Somebody should've added an <Ad> element to RSS long ago. Then filtering would be a no brainer.

Posted by Mark at 08:54 PM

July 03, 2005

Video capture, part XIX

transcode wolfs down CPU time like there's no tomorrow. Check out the drop off circled in yellow which is where transcode finished converting .avi to components for making an SVCD:

20050703.png

Even if you're burning standard DVDs, you still have to convert from DV to MPEG, and you'll probably use more CPU in a couple of hours of that than you'll use in weeks of gaming or years of desktop publishing.

What I don't get is that although a really usable, free home video application would be an absolute market maker for both Intel and AMD, you still have to buy a Mac if you're not a propeller head.

Posted by Mark at 07:43 AM | Comments (1)

July 02, 2005

Monopoly purchases immunity

Microsoft made a huge payout to Sun a while ago, and also handed out money to AOL and Gateway. Now according to The Register Microsoft's paying IBM to keep that company quiet as well.

Does somebody want to remind us all how this is going to be good for consumers?

On the other hand, you have to admit this looks a lot more intelligent than the old "Freedom to Innovate" campaign. BTW, if you want to give Microsoft more freedom to innovate, click here for information on how you can help. (Goodness. And I thought our Share campaign was embarrassing.)

Posted by Mark at 02:08 PM

July 01, 2005

Extensions

Ludo told me once about a bookmark synchronization extension to Firefox. You ftp your bookmarks over to your website and Firefox gets sync'd where you have the extension set up.

So I'm finally getting around to having a look at that. And found another extension, Auto Copy: "Select text and it's automaticaly copied to the clipboard. Like Trillian or mIrc"

Hmm. I thought that was already normal behavior in X. But I guess some windowing systems still don't do that. I tried to do some work on a Windows system a few months ago. It was horrible. Not only does select-paste not work, but the command line support is terrible. None of the software you want is installed, either.

Posted by Mark at 09:18 PM

June 10, 2005

User friendly

In the evening sometimes I make the mistake of letting my wife and children use the computer on my desk. Almost everything they do is done in a web browser. Occasionally Tim plays games that aren't browser-based, but that's rare.

Inevitably I get to put down the book I'm reading and play onsite support person.

Windows and windowing systems are too complicated. Flash errors are too complicated. The whole browsing model is tough. Popups and dialogs throw people for loops.

We assume so much. It's amazing anyone manages to use any of it.

Posted by Mark at 01:37 PM

June 07, 2005

Mouse, touchpad, & xorg

It seems to be xorg and this Toshiba Tecra. Both Solaris and Red Hat use xorg. No matter what I seem to do to the mouse section in xorg.conf, both the USB mouse and the touchpad work, but after a while the mouse pointer drifts off and starts rolling around the edge of the screen. Cannot figure it out.

Posted by Mark at 02:50 PM

May 15, 2005

Phishing spam

Spam comes in all the time. For the last year, lots of it has been phishing stuff such as what I got today in one of my mail accounts:

transfusable.gif

This had a link to emodule05.com, which according to whois, is registered to James Harris.

Getting spam didn't surprise me. What surprised me is that it is so quick and easy to find a name associated with the domain used for the spam scheme. I didn't realize phishing was legal. Can't you get in trouble for posting schemes to steal people's personal information?

Ah... okay, there are a few levels of indirection behind this... James Harris probably doesn't exist, any more than LeiMomi01 Design, tom.com, Wyith Ltd, etc.

Then you look these guys up and find a bunch of pages fighting spammers. Lots of bottom feeders out there.

Posted by Mark at 02:35 PM | Comments (1)

May 06, 2005

Usability

After Matt suggested perhaps Fedora might manage to handle the wireless device easily in the laptop on which I installed Solaris 10, I decided to try Fedora Core 3 in the disk space left over.

Two usability problems have kept me from using the laptop.

  1. Fedora installed Grub such that Solaris 10 won't boot.
  2. Fedora's support for the USB mouse that came with the laptop is flaky; After a minute or so the mouse pointer drifts off to the corners of the screen and I lose control over it.

Funny how those little configuration problems render the all software on the machine basically useless. Even funnier, I quit using the laptop because of them, even though I'm sure they're easy to fix.

Posted by Mark at 09:49 AM

April 12, 2005

Limits

I'm reading a rant list on C|Net called Technology's 10 most inexcusable failures. For one of the items, David Berlind writes:

Imagine if 10 years ago software vendors set themselves on a course to turn error messages into interactive software-repair assistants. Using the error dialog box, you would then be able to catalog the error in a log of your choice, forward it to some central repository (either corporate or with the vendor) and generate a trouble ticket for your support staff. Better yet, maybe you would be able to repair the problem with one click.

I'm imagining the situation right now. New managers working for the vendors would still be promising that software their predecessors promised to deliver 9 1/2 years ago, and junior developers hired to place those fired for not finishing would be cursing their managers' promises as code bloated exponentially with error messages in a vain attempt to account for unexpected conditions the software couldn't account for rendered even simple switches thousands of lines long.

Eventually, people would settle on dumping the entire problem in the hands of even more junior support personnel by automatically generating a contact-product-support URL -- i.e. one click -- for anything stronger than a warning. Encouraged by optimistic though incomplete support readiness training, crowds of customer facing youngers would go cynical overnight, confronted one call after another both with bitter IT staff suspecting unfair tactics on the part of product developers and angered by the wait on the phone, and also with second level support utterly at sea with a body of code devoted almost entirely to message logging, no one in the original development team having stuck around after management knotted the "Human readable messages/One-click resolution" must have millstone around their collective necks.

Posted by Mark at 09:11 PM

April 10, 2005

Building wine, part II

The build with wine-20050310 went fine, but I had to install the RPM with --nodeps. The build was from a source RPM and although everything worked with my earlier version of glibc, the spec file said there was a dependency on a later version.

Tibili appears to run fine under wine. I ran it with the default configuration file. The only additional step -- don't know if it was necessary -- was to put symbolic links to the CD drives.

Then I launched Tibili:

$ wine /mnt/cdrom/Tibili.exe

It came up in 640x480, but Emma and Tim played with it anyway. Emma played for about an hour. Tim then bugged her until she quit and left him the computer. He played to the end of the game.

Posted by Mark at 10:10 AM

April 09, 2005

Building wine

A long time ago I installed dosemu on Linux. Now I'm building wine. It's taking a while.

My hope is to be able to run J'apprends à lire avec Tibili under wine for Emma. She wants to use the program, and the 9-year-old computer in Tim's room is starting to flake out when reading CDs. I'd find a new CD reader, but am not sure it would work with Windows 95, which is the system on that old machine.

We'll see.

Posted by Mark at 10:12 PM

April 08, 2005

Laziness, part III

Solaris 10 runs fine on the laptop. One thing I notice right away is how easy it is to work here at Sun as a user without having to remember much at all about Solaris system administration, like how everything gets automounted...

One thing that strikes me right away is how nice it looks, how much the desktop is like a regular desktop. Yet underneath you have all the power and stability of a Solaris system. Very cool.

Posted by Mark at 04:07 PM

March 16, 2005

Case sensitive

I understand why man(1) needs case-sensitivity for arguments. But why can't we have a flag to turn that off?

Posted by Mark at 03:47 PM

March 15, 2005

Taxes online, part II

So I cheated, downloaded the library, sewed things back up, and Nathalie did our French taxes online... but " votre demande n'a pas pu aboutir. "

They've actually posted a note this year telling people to try signing their tax returns between 3:00-4:00 am to avoid the rush.

Posted by Mark at 09:44 PM

March 14, 2005

Niche news

C|Net's running a 2-page story on a forthcoming update to gcc. They consider it their top headline right now.

Is news just very slow? Has the general nerdiness in the population risen?

Since when is an as-yet-unreleased optimization of a C compiler a top story?

Posted by Mark at 09:12 PM

March 11, 2005

Maraboutage

Jean Véronis blogged a pretty good one the other day.

If you read French, check out the new docs.sun.com or Mark's Blog.

Posted by Mark at 03:26 PM

March 10, 2005

What is this?, part II

More spam bouncing back from my ISP. I got this image and message:

spam-20050310.gif

Letter build kind green, have. Cover, follow, learn. Winter too have. Man seem, doctor engine. High way develop under family. Life never way capital. Don't most correct window, black push. Began, bird finish they. Knew animal, began. Yet girl list fair hard. But inch radio hole blue. On fly, six voice. Little, special possible, stood see cent.

Here's the technical contact for the spammer's site, which was JJKAMAHIBD.COM:

Technical Contact:
    Potapova, Valeriya  vapotapova@mail.ru
    Donskaya 3 kv. 75
    Dushanbe, 676 127221
    RU
    +7.9262375937

Posted by Mark at 10:56 PM

March 09, 2005

Where they look

In his review of an article on how people read Google results pages, Darren Barefoot writes that he, "Never, ever look[s] at the ads."

On the one hand, the search in that image was for "digital camera" cheapest, so scanning the ads makes a bit more sense than when you search for "Directory Server 5" nsslapd-cachememsize. (Huh, that set of results came up ad free. Nobody wants to sell me extra memory?)

On the other hand, as Carole explained, those ads aren't for us. There are lots of people out there who still literally see the formatted content of a web page without preinterpreting it. My guess is that results of a study on people who ever think about website design or even just write web pages would show a different scan pattern.

Posted by Mark at 09:35 PM

March 06, 2005

What is this?

Although I cannot find the email address I got from my ISP through Google, I get bounces roughly weekly that show somebody's using it. But for what? An excerpt of one of the bounced mails:

Century office, down. Use pass grew. Minute yellow held if. Girl
will finger low plane friend. Even tire early it, after. Group
air where, written this look. Vowel, their here. Tell farm any,
other colony, practice. On can job.
--
Phone: 348-499-6730
Mobile: 628-542-9681

Then there's attached images that aren't images when you base64 decode them. Here's the base64 encoded "image" called reached6.gif:

R0lGODlhTAEdAJEAAAAAzMwAAAAAAP///yH5BAAAAAAALAAAAABMAR0AAAL/nI+py+0Po5y0
2otTyLz7D3aCYIwluZihtq3TiLqQarXyfdEuDOOOCowldAziIWAzIJVJn8LoHAolyCpryahq
bdtmVCoLfovTAdRcfqYH2KN3jDiP5dlNi3t/L7Bttv5LhyF2YMIjJdZzQsQzWKjzmHhSx8Lk
ZoXQVXm5RRZpSEgSeSjkSDOItndH6QbR1tcHilb2iSoax3i7aEuL2qsYK1mL22sa03haHOyr
4IqXylrpp8WsB6nMmIacTGycZtX0+gfth/mWaM3NC+yrTQi8jR5spPtOinLKfd3d0DxeHh0N
1io16/I1QIfw3r5UsMI9aMgphT19yiTeqoiPoryJ//UI5vqUEKNCRbTk9AOYiRw0gSk9LssI
pqNMYQ8vqQQ4yZ/AlwqH0QlZzxO2mTwrGoUCNOkaOksc3uSTBOrAnew4/rSKNVYgO1L9URt4
8yLRgi4N6ju3dGRBmPTKHhMLkwrXf0/z0IXodR1QMnCzStoaMKozujrN+Y3Jl+g8tT84anTk
Ed5bjWa34q27sus4qnsDtVNsUklYzaO9OEWs9EGQhab0MsaGD1JcdT2O8ZLtuA44PNMsqWQZ
says3J1yx2bau9xlnJZsjlq8rXEoe7qqE4eDHQ7V7GHWYOKOGLz48eRrlscRSDT59Ofbu18R
/D0I9ufpy7+PP7/+/fz7+9H/D2CAAg5IYIEG+gcAAAYk6ACDDzh4YIQSTghBghYe4CCEC2jI
AIcUfghigRpmqGAEHipwYogqrqjfiCWSiICFL5Y4wIkMypgAjgsqeOGOKbIIZJAS/AjjjkYW
iSKESs7oY49ICglllA0QyWSNNGLI45U5Xonkk1ZiKWWYYoLZYZUe6lhmjDPKWCWWbI4JZ5RU
Gvnllj1uyGWbam4ZZ59hukinl3WmCGidfO7pZ6JBFtolk0viieigegaqpaKWRvgmpWo+eWae
m9Jo452XjkpgAQA7

When base 64 decoded, the "image" contains bogus characters according to the Gimp. Gimp cannot read it.

The "images" are linked into the body of the message with stuff like:

<A href=3D"http://www.raster.nnjahglgdd.com/?p.XcrWpFVt0Tbdp00ccdb37">
<IMG alt=3D"" =
hspace=3D0=20
src=3D"cid:e100ccdb37@myIsp.domain" align=3Dbaseline=20
border=3D0></A>

The cid: seems to be some sort of Windows thing. Don't know what the 3D bits are.

Nnjahglgdd.com's home page:

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Frameset//EN"
   "_THE_LATEST_VERSION_/frameset.dtd">
<html>
<head>
</head>
<script>
var targetieSP2 = false;
</script>
<frameset frameborder=0 border=0 framespacing=0
 onload="if(targetieSP2) {IfSP2_load();}">
<frame 
 src="http://www.raster.nnjahglgdd.com/¤ê266/?affiliate_id=234227&campaign_id=0"
 name="list" marginwidth=10 marginheight=10 scrolling=Auto frameborder=no
 framespacing=0 >
</frameset>
</html>
$ whois nnjahglgdd.com
[Querying whois.internic.net]
[Redirected to whois.opensrs.net]
[Querying whois.opensrs.net]
[whois.opensrs.net]
Registrant:
 NA
 1753 Botany RD
 Banksmeadow, Sydney NSW 2019
 AU
 
 Domain name: NNJAHGLGDD.COM
 
 Administrative Contact:
    Platt, Mather  mather_platt@yahoo.com.au
    1753 Botany RD
    Banksmeadow, Sydney NSW 2019
    AU
    +61.294750668    Fax: +61.294750668
 
 Technical Contact:
    Platt, Mather  mather_platt@yahoo.com.au
    1753 Botany RD
    Banksmeadow, Sydney NSW 2019
    AU
    +61.294750668    Fax: +61.294750668
 
 
 
 Registrar of Record: TUCOWS, INC.
 Record last updated on 28-Feb-2005.
 Record expires on 28-Feb-2006.
 Record created on 28-Feb-2005.
 
 Domain servers in listed order:
    FIRST.XBCBCDEE.INFO
    SECOND.XBCBCDEE.INFO
    THIRD.XBCBCDEE.INFO
 
 
 Domain status: ACTIVE

That email address brings me to a Japanese blog full of blog spammer stuff, http://swatteam.blog.ocn.ne.jp/my_weblog/2005/03/.

Explanations?

Posted by Mark at 08:12 AM

March 02, 2005

Design documents

Jack W. Reeves wrote a paper some years ago about software design, defining the design to be the source code. He recently wrote his own review of the discussion and criticism of his ideas, in which he states what he means by a design document:

When the document is detailed enough, complete enough, and unambiguous enough that it can be interpreted mechanistically, whether by a computer or by an assembly line worker, then you have a design document. If it still requires creative human interpretation, then you dont.

That's still a refreshing view on what purpose source code serves, and suggests why you might need to revise, revise, revise even after you get something that seems to work according to the requirements identified. It also makes me think that most software design is beyond the comprehension of anybody not working on the software. It explains as well why good software design is hard.

Posted by Mark at 02:39 PM

February 27, 2005

Runner's log

I found Run at Sourceforge.net, and have been trying to set it up at Mcraig.org. Cannot get the access straightened out.

Until security is easier to understand for the administrator than for the crack, the outlook is bleak. In other words, the outlook is bleak and will probably remain so until beyond the end of time.

Posted by Mark at 09:39 AM

February 22, 2005

Full text, part II

As Tilly alludes in the comments for my entry on this topic yesterday, Jean Véronis's Technologies du Langage blog covers the problem in more detail. In a recent entry Véronis demonstrates some amusing bugs in how Google references web pages, but suggest the crux of the matter lies deeper:

Or, la grande faiblesse de Google est justement son manque de chercheurs dans le domaine du traitement des langues. L'analyse des domaines de compétence de ses chercheurs travers leur CVs et leurs publications fait apparatre une absence quasi-totale d'expertise dans ce domaine. Une telle expertise existe chez les développeurs de petits moteurs (notamment en France), mais les petits David semblent bien faibles par rapport au grand Googliath.

(Source: Référencement: Drôlement verni !)

Interesting point, and a good reason to keep alternative engines alive. But going back to Jeanneney's plea, in getting the full text on the web are the Davids in competition with the Googliath? After it's out there, will it really matter that Google makes a hash of the work of referencing the content?

Granted there won't necessarily be buckets of stock options and $3M bonuses, 1000 new millionaires, when the software editors do quality work rather than focus first on advertising revenue. (Unless David sells out to Goliath in return for quickly vesting options.) But fast bucks don't seem to be the goal. In the case brought forward by Jeanneney, we're talking about availability of texts selected by European librarians. The problems are separate. The solutions may be as well.

Posted by Mark at 09:44 PM

February 19, 2005

More Gmail

Something I only noticed yesterday, and am reminded of by this C|Net article, Gmail no longer provides only 6 invitations at a time, but 50.

There was some hubbub about the safety of what you store there. I consider it another virtual desk surface, browser-accessible, full-text-searchable, and backed up by someone else. As Scott McNealy once said before Sept. 11 and before everybody recognized Identity Management as a huge market:

You have no privacy. Get over it.

What you do have is anonymity. Nobody cares about you like you care about yourself. Not even your mother or John Negroponte.

Let me know if you want 1 GB on their server.

Posted by Mark at 09:17 AM

Slower and slower, part II

mcraig.org and in particular the CGIs that make up MovableType seems slower and slower. Not sure whether they're doing maintenance when I create entries. When I rebuild, it takes a while. Perhaps there's something that gets longer to do the more entries you have.

Posted by Mark at 09:09 AM

February 16, 2005

DITA in public review

The Committee Draft for the OASIS Darwin Information Typing Architecture (DITA) 1.0 is available for public review. As mentioned in email:

Comments may be submitted to the TC by any person via a web form found on the TC's web page at http://www.oasis-open.org/committees/dita.

Standardized modular documentation. Neat idea. Looking forward to the applications.

Posted by Mark at 08:23 AM

February 14, 2005

Moving MovableType, part IV

I've decided to allow comments without and use mt-close.cgi, nofollow.pl, and the new interface in MovableType 3.x to combat comment spam. If you see something telling you to wait for approval, let me know.

Posted by Mark at 07:53 AM | Comments (2)

February 13, 2005

Video capture, part XIII

Video capture with year-old hardware, Linux IEEE 1394 modules, and a 2.4.21 kernel makes for a CPU-intensive job.

sysmonitor20050213.png

It's a drag for the user, since you don't see anything happening except the system slowing down.

Posted by Mark at 08:18 PM

February 11, 2005

Moving MovableType, part III

After moving the blog, I regenerated the whole thing from the exported content. Links to articles in this blog came up as http://mcraig.org/mark/archives/000690.html.

Should be no problem, right? Theoretically that's correct. But a couple of times I had to delete articles, leaving holes in the numbering. I don't think MovableType left holes in the numbering when it regenerated the blog from exported content.

So W3C-LinkChecker online found only valid links, except for the ones to Amazon.com which didn't support getting just the head and so gave 405s. But some of those links might have been misdirected, semantically speaking. Cannot think of a way to let the machine check for those kinds of errors.

So if you go trolling through the archives and get a link to apples when you're sure it should be oranges, let me know.

Posted by Mark at 03:58 PM

Moving MovableType, part II

Andy and Dana had posted comments I wasn't seeing right away. MovableType 3.15 is configured by default to require that the blog owner approve comments before they appear.

It's a good idea, but not all of us read the release notes or the changelog very carefully. (Less carefully than we write them, anyway.)

Posted by Mark at 03:09 PM | Comments (1)

February 04, 2005

TypeMover

Made a backup with TypeMover last night, which is great.

The only thing missing if I want to move the blog are dozens of photos.

Posted by Mark at 03:40 PM

February 03, 2005

MovableType 3.15

Yet another upgrade, this one to take advantage of a security fix and generally to catch up with the Joneses after feeling guilty about my laziness.

My brother Matt and the geeks he hacks with are apparently constantly on the bleeding edge of Gentoo, so I'm pretending to be an early adopter by relation. (Note: as suggested, I also reinstalled a couple of workstations in the lab with Solaris 10.)

Ludo reminded me today at work that my problem last night was not reloading the page when the CSS got stuck in a cache somewhere. The 3.x CSS does make MT look cleaner than the old version.

Posted by Mark at 09:19 PM

MovableType 3.1, part II

Looks like whatever CSS issue I was having yesterday has worked itself out, as if by magic. The pages looks fine in Firefox 1.0 here at work.

Posted by Mark at 02:34 PM

February 02, 2005

MovableType 3.1

Just upgraded to MovableType 3.1.

The upgrade went smoothly, but there's something horribly amiss with the CSS for this page for editing the blog in Firefox. Everything is stuck inside the column on the left edge of the page.

Posted by Mark at 10:40 PM

February 01, 2005

Laziness

Have become very lazy about software upgrades lately.

Matt, my brother, is eventually going to move this domain somewhere. I'm too lazy even to have untarred the new version of MovableType I downloaded last month. Need to have a look.

I'd also been too lazy to upgrade to Thunderbird 1.0 and Firefox 1.0. That's like watching commercials because you're too lazy to push the button on the remote. Finally got around to that.

Ludo reminded us to upgrade to Solaris 10, which he observes runs faster on his workstation than the previous version he had. I've only installed Solaris 10 on an old PC I rarely use, and even then it was mainly out of laziness: easier to install the machine fresh than to patch it before installing some software I needed to test.

On a laptop I use, I've been running a beta version of our iWork Java Desktop System ever since the beta eval started.

On this system at home, I'm still running Red Hat 9 with a 2.4.21 kernel.

I have an old Vaio laptop still running Windows 98 (for device support to transfer files from a Sony memory stick that I couldn't easily get working with Gentoo last time I tried).

I've even been running Directory Server 5.2 patch 2 for months and months on my old workstation that's now in the lab, yet I doc the new versions for that.

Maybe it's not pure laziness. Part of the problem is that the existing stuff actually works pretty well.

Posted by Mark at 10:07 PM

January 28, 2005

Schema repository, part XI

Hacking away at man page generation this afternoon, I ended up needing the URL to the server where I'd stored my repository. I was iterating through a collection of schema objects to get olinks to their man pages, and needed to read the actual object content to get the prefix (a sort of short name):

        iter = objects.iterator();
        try {
            while (iter.hasNext()) {
                LDAPUrl objUrl = new LDAPUrl(baseUrl + (String)iter.next());
                SchemaObject obj = COLL.createSchemaObject(objUrl);
                list += getOlink(obj.getPrefix(), obj.getSuffix());
                if (iter.hasNext()) list += ", ";
            }
        } catch(Exception e) {
            e.printStackTrace();
        }

So don't tell anybody: I just hard coded it in there and left myself a //TODO! comment.

Roughly when I typed e.printStackTrace(); I realized that the core Java code, which has grown to 2053 lines, is crying out for me to go back to the design I scribbled one day on my white board.

Revise, revise, revise. Isn't that what the real writers say?

Posted by Mark at 09:28 PM

January 25, 2005

WordPress

Ludo's checking out WordPress as an easy-to-configure, open-source blog platform, all in PHP with an SQL backend. Another something to look at when I finish playing with Netbeans.

Posted by Mark at 06:22 PM

January 21, 2005

Aggregation

Ludo put together an aggregation of blogs at work today using Planet. Pick a topic and Presto! instant journal.

Or almost. If you look at Technorati tags right now, Fri Jan 21 21:33:54 CET 2005, the biggest English word I see is "General" with "Main Page," "Diary," and "News" not far behind. "Uncategorized," "Current Affairs," "Personal," "Blog," "Weblog" are right up there. Not to mention "Actualités," "Algemeen," "Allgemein(es)," "Miscellaneous," even "Stuff." Still need a human being to pull worthwhile aggregations together.

"There is no cheap metadata."

Posted by Mark at 09:42 PM

January 20, 2005

rel="nofollow"

Google's no longer giving pagerank credit to links marked up with a rel="nofollow" attribute. You add this to links in comments from folks not registered.

The idea is to leave no incentive for blogspammers to post a bunch of links back to client sites to raise the rank of those sites. I just installed MovableType's plugins to add the functionality.

Posted by Mark at 09:00 PM | Comments (1)

January 18, 2005

Netbeans 4

Started playing with Netbeans 4 today, as I wanted to use it to refactor the code I'd hacked together for generating man pages from my proto schema repository. So far it looks promising. (Netbeans, not my code)

Definitely nicer on a 21" screen at 1600x1200 than a laptop screen at 1280x960. Plenty of little windows indicating all sorts of things.

Posted by Mark at 09:48 PM

January 15, 2005

Intellectual property, part X

Yesterday Matt noticed a US patent granted for the idea of a hash table. He'd been reading a Slashdot story on file sharers putting themselves in legal danger by abusing intellectual property rights.

We considered writing up a joint invention disclosure on a process that we've reduced to practice, which involves taking air into the lungs, where oxygen is traded for carbon dioxide and trace waste gases before the air is expelled. Then Matt worried there may be prior art on that one, so we got lazy and went back to work.

Posted by Mark at 07:19 AM

January 06, 2005

Solaris libre, part II

C|Net has a story stating that the Open Source Initiative's license approval committee gave the okay to Sun's community development and distribution license.

I don't know any more than you do, but that makes it look like we're preparing open source Solaris in earnest. I wonder what that means for other software from Sun.

Posted by Mark at 09:36 AM

December 29, 2004

Gmail, part II

Google mail seems to have offered me 6 more invitations. Let me know if you're interested. First come, first served.

Posted by Mark at 07:12 AM

December 04, 2004

Buying a flight

Looking for fares to fly the family to Chicago and back during summer vacation, I've tried expedia.com, opodo.fr, and now directly the Air France site.

Expedia.com won't deliver paper tickets outside the US. You find that out near the end where they have all the details you entered and just want your credit card number.

Opodo.fr -- "appartient Air France, Aer Lingus, Alitalia, Austrian Airlines, British Airways, Finnair, Iberia, KLM, Lufthansa et Amadeus" -- recommends starting with the train to Paris. Opodo.fr's cheapest finds are about 50 euros more expensive than what I get directly at the Airfrance.fr site!

Is it the software alone? Opodo.fr probably has to look up flights that are in some shared place, whereas Airfrance.fr can look directly in the Air France backend. Still, I wonder where I can get a real price comparison.

Then there's the problem of layovers. How short is too short? When they propose a 50 minute layover although my query includes 3 children under 12, and you have to take your shoes off and get scanned for most international travel, why didn't they think to let you set a minimum plane-change time? Or, better yet, why didn't they do that for me?

How well do they think they're doing? This software stuff is hard, isn't it.

Incidentally, I do have a real problem with the Airfrance.fr site: Your party can be no larger than 4 max., otherwise you cannot reserve your flight. Too bad for people with families of 5.

Posted by Mark at 02:48 PM | Comments (2)

December 01, 2004

cocolicious.us

Matt seems to have managed to setup cocolicious.us at home using dynamic DNS. No web server yet, but I've already received email from him.

I'd run a site from my basement, but we have too many power cuts in my neighborhood.

Posted by Mark at 09:47 PM

November 29, 2004

Video capture, part V

On my system at home is an old version of cdrecord without support for CUE files. So I'm burning an SVCD with cdrdao.

It doesn't directly support my CD-RW drive, so I'm using the generic driver.

The article recommending I try cdrdao also recommends I burn it slowly for maximum hardware compatibility. I wonder if this one will work in our livingroom DVD player.

Posted by Mark at 09:54 PM

November 28, 2004

Video capture, part IV

After all that capturing and transcoding, I now have a bunch of .mpg files. When I put them together into a single SVCD file, the result takes up 840 MB. I'll have to figure out how to cut the SVCD in half. Good thing Kino's default was one sequence per .avi file to start with on export.

Posted by Mark at 10:46 PM

Video capture, part III

Moving digital video to an SVCD is currently a slow process, even on a computer with components no older than a couple of years. The slow bits, such as transcoding from NTSC in 720x480 to SVCD format in 480x480 runs at no more than about half real speed on this AMD 2400+ system.

If I try a game of Quake at the same time -- the demo version runs on Linux -- the game's almost too choppy to play, and the transcoding slows to about 1/3 real speed.

Posted by Mark at 10:05 PM

November 20, 2004

Changing hosting

Matt's thinking of changing hosting providers for this site. Makes sense, but it may take me a little while to get everything fixed up to blog again. Not sure yet when he's going to make the change.

Posted by Mark at 09:06 PM

October 29, 2004

Broken windows

Something strange happened on the Vaio laptop. Every time I try to copy paste from the memory stick, Windows 98 crashes and reboots. Had to move images in place with pscp.

Posted by Mark at 07:08 PM

October 15, 2004

Google desktop

Check out this idea: Google Desktop.

It's almost enough to make you want to run Windows. A real shame they don't support useful operating systems, yet.

Posted by Mark at 02:40 PM

October 06, 2004

Fear of StarOffice?

The Register has Steve Ballmer acknowledging that StarOffice, and by extension OpenOffice as well, exists in competition with Microsoft's big cash cow. There must be some big customers complaining they cannot see why they should pay through the nose for office productivity (sic) software.

Ballmer's quoted as disparaging StarOffice because, "Its not compatible with Microsoft Office and its missing key applications like Outlook." There probably are some things that aren't compatible. That's perhaps because StarOffice and OpenOffice use a public standard XML format to store documents, whereas other folks rely on proprietary formats to get your arm and leg.

It's not entirely clear why Outlook would be considered a key application, except by people who write email viruses. You definitely don't need it to send mail, manage your contacts, keep your calendar up to date, or any of that sort of thing. Perhaps it's key to keeping users captive and to propping the stock price up.

Maybe Ballmer's finally learning from Andy Grove. Someone once suggested that Intel went out of its way to prevent potential competitors from dying off completely, because Grove realized Intel's monopoly would be unambiguously identified as such had they no viable competition at all.

Someday we'll learn about the hush funding wired from Redmond to a student bank account in Finland back in the early 90s.

Posted by Mark at 09:28 PM

September 29, 2004

Skips in RSS?

Huh? Since I updated the configuration to publish the first 500, instead of the first 40, words of each article, Movable Type seems to have resyndicated everything on the top page of my blog.

The strange behavior? When I look at the results through bloglines.com, this entry is missing.

Posted by Mark at 12:09 AM

September 18, 2004

LDAP schema repository. part II

After a long hiatus, I started playing again with the schema repository idea.

The object of my play at this point is to pull the schema definitions, which are attributes values, off cn=schema from a directory, and to have objects for handling definitions in a Java program. JNDI seems to have most of what I need, but I'm also using the java.util.regex so I have the list with which to call lookup().

Maybe I've missed something, but it looks like the tools JNDI provides are great for handling the schema definitions of a particular object or attribute, but not so natural when you want to manage the schema itself, independent of any instantiated objects.

Posted by Mark at 02:28 PM

September 17, 2004

Ashamed to back out

Okay, I got tired of how little I really understand about kernel configuration, so I dropped Windows 98 back on the Vaio.

emma20040917.jpg

Emma doesn't care.

Posted by Mark at 04:20 PM | Comments (1)

Gentoo, part II

With the previous Gentoo install, I built a 2.4.26 kernel. Couldn't get the support for my Sony memory stick working.

From a quick search on Google, it looked like people having luck with memory stick support were using 2.6 kernels. So I tried Gentoo with a 2.6 kernel. No dice.

I'm going to have to eat my words pretty soon if I don't get this one solved. (BTW, JDS does not seem to want to be installed on my old Vaio.)

Posted by Mark at 02:07 PM

September 15, 2004

Gentoo

"And now for something completely different..."

My brother Matt suggested I try Gentoo Linux. Why not?

Gentoo is not for marketing dudes, but it comes with nice documentation for newbies. Don't try it unless you've grokked Zen and the Art of Motorcycle Maintenance, however.

Good thing Google is there when you need to learn something. The first bit I had to learn was passing ide2=0x180,0x386 to the kernel on the install CD so my Vaio could read the CD-ROM after the system settings replaced the BIOS. Then at the end, I'd forgotton to set a root password. This was the first time I've edited files with echo and sed because no interactive text editor was available.

Posted by Mark at 05:06 PM

September 05, 2004

Dosemu

C|Net appears to have run yet another story Friday about Linux posing a threat to Microsoft. It's not clear how Linux poses a serious threat to Microsoft. Perhaps if you run a monopoly you need to pretend you're in danger from the competition. The more your friends in the media publish the story, the more true it sounds to people who only have a dim understanding of what's going on.

It's kind of like the folks in Washington saying the US is in danger of being invaded by Iraq, and the press falling into line. Or I take that back. Microsoft will probably stop short of dropping huge bombs approximately on people who write open source software.

Anyway, we still have two systems running Windows. Tim has the old Pentium 100 based PC running Windows 95. And I have an old Vaio laptop running Windows 98, because I was too lazy one afternoon to figure out whether Sun's Java Desktop System includes support for Sony compatible memory sticks. It probably does, but I had a couple of photos to upload, and took the easy route of reinstalling the laptop from the boot CDs. I figured certainly Sony would support the memory stick, and they do. It's kind of a drag to have the system crash on boot about 1/3 of the time, but I only use it every other week.

Tim occasionally runs Windows 95 to play games. His games include Adibou and Tibili. Under Linux, he also likes to play a Quake demo we downloaded, but that one would never run on the old system anyway.

Emma's computing is system independent. She plays Flashware at Barbie.com, and that's it for now.

Beyond the drivers for handling the memory stick, I have one lingering possible use for Windows, that of opening some old Cakewalk for Windows files I created in 1996. Any time I get nostalgic, I can use dosemu to play old shareware, like Duke Nukem 3D.

Interestingly Duke3D works better in dosemu than it ever did on my on the PC Tim now has. It even runs in 800x600 instead of 320x200. I wonder if somebody has updated that one to run it on a more recent underlying platform. Quake is humorless by comparison.

Posted by Mark at 06:12 PM

September 01, 2004

Gmail

Ludo invited me to use Gmail, so I thought I'd try it out.

More than another web-based email site, Gmail gives you Python access to a GB of web-accessible space.

$ df /gmailfs
Filesystem           1K-blocks      Used Available Use% Mounted on
gmailfs                1024000         0   1024000   0% /gmailfs

Interesting idea... I didn't see right away how to mount the file system through the firewall at work, however.

Posted by Mark at 09:30 PM

August 30, 2004

Comment spam, part II

Another small rash of comment spam happened today. Somebody posting from two different IP addresses posted 16 ads for a Viagra substitute called Cialis, advertised as producing a 36-hour erection.

I've not tracked the IP addresses, but my guess is that the IP banning in Movable Type is not going to cut it. Like most of us, my IP address at home, for example, changes every time I get a connection. Movable Type lets you ban individual addresses.

Perhaps I should try MT-Blacklist?

Movable Type, by the way, has an editor for handling multiple entries at once, but I don't see one for selecting and deleting multiple comments. Were they waiting to put that in the paid license version?

Posted by Mark at 08:10 AM

August 15, 2004

Bit torrents and big brother

I've come to the party very late, but have started to use Azureus to get and send bit torrents. I haven't tried to figure out how it works, but seems to function peer-to-peer for exchange of large downloadables.

There's a lot of shareable bits out there that people seem to want to diffuse more than they want to copyright and prevent you from checking out until you pay. For example, I've downloaded a Grateful Dead and Trey Anastasio shows from torrents found at nugs.net, which says it distributes only free music. I'm downloading a couple of VCD files of a talk by Chomsky last year concerning the world after after the invasion of Iraq. You can probably get GPL-ware this way, though I haven't looked too hard.

Trouble is, if you go Googling for a bit torrent, it's not necessarily clear what's shared legitimately and what's shared illegally. The University of California at Riverside has posted guidelines for example explaining to students how to comply with the DMCA:

Very simply, do not download or distribute copyrighted materials without appropriate permissions.

In other words, it's up to you to know. Furthermore, Chuck Rowley reacted in what seems like a very normal way for a sys admin to react to something like the DMCA:

In compliance with the 1998 Digital Millennium Copyright Act (DMCA) and the University of California Guidelines for Compliance with the Online Service Provider Provisions of the Digital Millennium Copyright Act, UCR expeditiously takes action when notified of potential DMCA violations from sites located on the campus network. All of these incidents are referred to various campus officials and appropriate actions are taken to stop unauthorized downloading or distribution of copyrighted materials.

What other choice do you have if you want to keep your job?

So these "various campus officials" probably have overflowing inboxes, because all file sharing potentially violates copyright. It's a shame that instead of DMCA there's not just an RFC or two on labelling downloads cheaply as copyright protected, copyleft or whatever. Download and upload apps could then verify for themselves whether the material is protected, and could ask for confirmation before performing operations whenever it looks like the material is not marked as freely sharable. You could even have a protocol for handling it, which the military probably already has, anyway.

Chuck Rowley at UCR has the right idea when he writes:

Of course, there are legitimate applications of file-sharing software, and discussion as well as research on such peer-to-peer software is expanding rapidly in the academic community. We will ensure that such inquiry, as well as the legal use of peer-to-peer software, remains unimpeded at UCR.

It's a shame that our overburdened and probably for the most part underconcernced elected officials did not find a way to make it a snap for everyone to do the right thing, rather than vote in the DMCA, a 94-page legal document that covers both downloads and vessel hull designs, and includes such dubious provisions as the one disallowing you to "reverse engineer" software (whatever that means) for any other purpose than figuring out how to interoperate with closed software. Presumably any documentation I do without asking beforehand is therefore in violation of DMCA. Good grief!

Posted by Mark at 01:51 PM

July 24, 2004

Impossible to back up

Once you start playing with digital video, you start having comparatively large volumes of data you'd like to back up. Theoretically, you can put your work on master tapes. Theoretically I write because I have not yet managed to get that to work.

So at 1 GB per 4.5 minutes of video, things quickly got to the point where I cannot back up my home dir anywhere but another hard drive. (No money for a DVD burner right now.)

When reinstalling everything, I have this mental picture of me with a limited number of vessels pouring precious liquid from one to another, hoping not to find any cracks.

Posted by Mark at 09:16 AM

July 23, 2004

More on security

This stuff sure seems complicated to me.

Latest challenge: understand SASL enough to write some good examples.

Posted by Mark at 07:06 PM

July 21, 2004

Secure with head in sand

Back when my brother and I started playing with gpg, I hardly understood what the heck was going on. Assuming I'm roughly as dumb as the average network user, there must be a lot of us folk out here cluelessly attempting to secure our data.

On a slightly more frightening note, I'm now opening ports in my firewall for work, probably way too widely. Why don't I understand? Partly just laziness. Partly the situation is bad-habit forming like drinking and driving, in that you often get away with it until you have an accident, feeling secure since you cannot see the danger with your head buried in the sand.

Security software works fine "when used as directed," PKD wrote in Ubik. Trouble is, the directions are complicated enough that everybody except Luke just sucks it and sees...

Posted by Mark at 10:50 PM

At long last!

Finally, I found the problem with the VPN client. My firewall wasn't letting in the traffic from the server side. Shouldn't that have been obvious from the logs?

That, believe it or not, was in the Cisco docs, although the details don't quite apply in my case, which is perhaps partly why I skipped it on the first pass.

Posted by Mark at 12:02 AM

July 20, 2004

Now with logging

Okay, I am the end luser here. I've got logging turned on; the doc doesn't help. Here is an excerpt what I get to try to decipher:

4      21:41:24.902  07/20/2004  Sev=Info/4     CM/0x43100002^M
Begin connection process
                                                                                
5      21:41:24.902  07/20/2004  Sev=Warning/2  CVPND/0x83400011^M
Error -28 sending packet. Dst Addr: 0xC0A800FF, Src Addr: 0xC0A80063 (DRVIFACE:1236).
                                                                                
6      21:41:24.916  07/20/2004  Sev=Info/4     CM/0x43100004^M
Establish secure connection using Ethernet
                                                                                
7      21:41:24.916  07/20/2004  Sev=Info/4     CM/0x43100024^M
Attempt connection with server "vpn.example.com"
                                                                                
8      21:41:25.137  07/20/2004  Sev=Info/4     CVPND/0x43400019^M
Privilege Separation: binding to port: (500).
                                                                                
9      21:41:25.137  07/20/2004  Sev=Info/4     CVPND/0x43400019^M
Privilege Separation: binding to port: (4500).
                                                                                
10     21:41:25.137  07/20/2004  Sev=Info/6     IKE/0x4300003B^M
Attempting to establish a connection with aaa.bbb.ccc.ddd.
                                                                                
11     21:41:25.207  07/20/2004  Sev=Info/4     IKE/0x43000013^M
SENDING >>> ISAKMP OAK AG (SA, KE, NON, ID, VID(Xauth), VID(dpd), VID(Nat-T), VID(Frag), VID(Unity)) to aaa.bbb.ccc.ddd
                                                                                
12     21:41:25.208  07/20/2004  Sev=Info/4     IPSEC/0x43700008^M
IPSec driver successfully started
                                                                                
13     21:41:25.208  07/20/2004  Sev=Info/4     IPSEC/0x43700014^M
Deleted all keys
                                                                                
14     21:41:30.242  07/20/2004  Sev=Info/4     IKE/0x43000017^M
Marking IKE SA for deletion  (I_Cookie=2064A5FC4423F0EF R_Cookie=0000000000000000) reason = DEL_REASON_PEER_NOT_RESPONDING
                                                                                
15     21:41:30.781  07/20/2004  Sev=Info/4     IKE/0x4300004A^M
Discarding IKE SA negotiation (I_Cookie=2064A5FC4423F0EF R_Cookie=0000000000000000) reason = DEL_REASON_PEER_NOT_RESPONDING
                                                                                
16     21:41:30.781  07/20/2004  Sev=Info/4     CM/0x43100014^M
Unable to establish Phase 1 SA with server "vpn.example.com" because of "DEL_REASON_PEER_NOT_RESPONDING"
                                                                                
17     21:41:30.782  07/20/2004  Sev=Info/4     CM/0x43100011^M
Attempt connection with backup server "vpn2.example.com"

It's just so obvious! The when I google for DEL_REASON_PEER_NOT_RESPONDING, even the discussions are incomprehensible. No wonder Sun IT only wants to support this on platforms they control.

Posted by Mark at 10:26 PM

July 15, 2004

2.6.7, part III

When I thought I was debugging, I was in fact rendering the 2.4.21 kernel unbootable.

Wish I'd had as few problems as Norman Walsh.

Too discouraged to go any further today.

Posted by Mark at 08:51 PM

July 14, 2004

2.6.7, part II

Got tired of make install failing, so finally I just created an entry next to the other ones for grub and rebooted.

2.6.7 boots, but with some failures. X of course does not work, because of my on-board Nvidia graphics support. I even hosed my 2.4.21 nvidia.o impatiently letting the Nvidia module builder overwrite the old module. On-board Ethernet didn't work either, so there I was stuck without even web access.

By then I was flustered and had to calm down for a while before managing to think naturally and debug the problem. Guess it's time to read up on 2.6 before I try to go further.

Some of the dmesg makes good copy:

...
ehci_hcd 0000:00:02.2: irq 5, pci mem f881c000
ehci_hcd 0000:00:02.2: new USB bus registered, assigned bus number 1
irq 5: nobody cared!
Stack pointer is garbage, not printing trace
handlers:
[] (usb_hcd_irq+0x0/0x70)
Disabling IRQ #5
...

Why change the kernel, anyway? That's a good question. Maybe if you're not good enough to fix things that are broken, you naturally start trying to fix things that aren't.

Posted by Mark at 10:02 PM

2.6.7

Building the 2.6.7 kernel, I notice the changes -- no more make dep, modules are now .ko. I also notice that something's fishy with the ACARD SCSI module:

[root@lethe linux-2.6]# make install
make[1]: `arch/i386/kernel/asm-offsets.s' is up to date.
  CHK     include/linux/compile.h
Kernel: arch/i386/boot/bzImage is ready
sh /usr/src/linux-2.6.7/arch/i386/boot/install.sh 2.6.7 arch/i386/boot/bzImage System.map ""
No module atp870u found for kernel 2.6.7
mkinitrd failed
make[1]: *** [install] Error 1
make: *** [install] Error 2
[root@lethe linux-2.6]#  ls /lib/modules/2.6.7/kernel/drivers/scsi/atp870u.ko
/lib/modules/2.6.7/kernel/drivers/scsi/atp870u.ko
[root@lethe linux-2.6]#

Furthermore, during the make, this went by:

Kernel: arch/i386/boot/bzImage is ready
  Building modules, stage 2.
  MODPOST
  CC      drivers/ide/pci/amd74xx.mod.o
  LD [M]  drivers/ide/pci/amd74xx.ko
  CC      drivers/scsi/atp870u.mod.o
  LD [M]  drivers/scsi/atp870u.ko

Hmm...

Posted by Mark at 07:14 AM

July 07, 2004

Java GetOpt method

This afternoon I was looking around for a getopt() Java equivalent. Looking on Google with "I'm Feeling Lucky," and what do I run into?

netscape.ldap.util.GetOpt!

Of course I was playing with the examples for that SDK. I looked around gently for netscape.ldap.util.kitchenSink, but maybe that class isn't public.

Posted by Mark at 10:37 PM

videoconv.sh, part II

In videoconv.sh, I wrote, "It looks like perhaps something may be goofy in the options passed to mplex." Sure enough, my version of mplex has no -m option...

The script also has a hardcoded value for PAL speed video (25 fps) with the -F option getting 3 (value for NTSC: 4).

Posted by Mark at 10:26 PM

5 hits from US Military

mcraig.org got 8 hits from the Seychelles in June, but we also got 5 hits from domains allotted to the US Military.

A search for Chomsky brings up 5 hits in my blog. But 2 of them are dated July 2.

Maybe they're asking the oracle about potential bombing patterns in Syria?

Posted by Mark at 10:21 PM

Yikes, my mail!

I'm playing around with old software docs, in which it recommends you add something to your prefs.js to bypass having to get an SSL certificate. So I try that.

Ok. Mozilla seems to run fine. The example I'm hacking on doesn't work, but that's fine, too. It was written for Netscape Communicator and things have probably changed.

What I did was let applets access the network without authorization, so I go to turn that off. At the top of prefs.js is something that looks like a comment, a line starting with #. So I add a # in front of the line I added and restart Mozilla.

No mail. Nothing is configured. Getting your mail configured with the right filters and everything is of course something that only takes a couple of hours... if you can remember what you did, and have all the account details handy.

No error messages, all the stuff is still there in prefs.js.

Take away the #, and my mail conf comes back. Whew!

I guess I'll take that recommendation out of the docs when I update them...

Posted by Mark at 11:23 AM

July 06, 2004

videoconv.sh

In the ongoing saga of converting my .avi files to a video CD, I've started using a front-end to vcdimager called videoconv.sh, since vcdimager doesn't handle the conversion to mpeg, and that seems like it may be what I got wrong before.

Unfortunately, videoconv.sh seems to have a little bug that I only noticed 90 minutes into the conversion. In an attempt to cleanup potentially large temporary files, the script deletes them inside the for loop that processes the .avi files, right after the temporary files are supposed to be used. Trouble is, the temporary files are not used, and so are deleted before they get processed into something else. Since lav2yuv in quiet mode generates 1 line of output per frame (29.97 frames/second in NTSC) processed, you can hardly scroll back to see what happened unless you were prepared in advance to expect errors. I'm too trusting ;-)

It looks like perhaps something may be goofy in the options passed to mplex. Another step in the saga...

Posted by Mark at 10:26 AM

July 05, 2004

What's a blog, Mark?

Dad wrote:
> By the way, what's a blog?

An electronic diary. Tim Bray, a sort of internet architect who works at Sun, told us the other day in a presentation that the number of blogs out there has been doubling every 4-6 weeks lately. When I was last out in California, John Fowler, our software CTO, told us Sun hired Tim to investigate blogging and how Sun should get involved. To kick it all off, Sun published a website of employee blogs: "Welcome to Blogs.sun.com! This space is accessible to any Sun employee to write about anything."

What seems to be at the root of that is something called RSS, Really Simple Syndication, a protocol for summarizing articles from diaries, journals, online news, really anything that's regularly published.

Even our ponytailed COO, Jonathan Schwartz, has started a blog.

So we can now smell the tulip (bulb) odor surrounding this phase of the phenomenon. Blogging's bound to go out of fashion soon, kind of like LinkedIn or orkut. Yet, you can imagine that rolling journals like server logs together using RSS or a similar protocol may well have a future.

One advantage blogging has over email is that I can write this up, send you a permalink, and it persists out there on the web as a bit of writing for which you can Google, and to which you can link. Disadvantage, as they say at Slashdot: it goes down on your permanent record.

Posted by Mark at 07:57 AM

July 02, 2004

Seems mostly fixed

It looks like I've mostly fixed the mistake I made adding an extra / to my rm command the other day.

Rather than reinstall over my existing install of Red Hat 9, I found a system on which I could install all the packages, then copied over the missing files. I seem to have found most of it. Things appear to work at first view.

Posted by Mark at 10:13 PM

June 30, 2004

NTFS blocks Linux

When you're stuck and don't have enough space, it's a drag to be reminded that Microsoft does what it can to keep Linux developers out of the game. The NTFS module on my JDS install can only read, not write to, the Windows system partition, which has 4.5 GB of therefore unusable free space:

lethe:~/redhat # df -h 
Filesystem            Size  Used Avail Use% Mounted on
/dev/hda8             8.4G  4.5G  3.9G  54% /
/dev/hda6             7.6M  6.5M  790K  90% /boot
/dev/hda1             7.5G  3.1G  4.5G  41% /windows/C
/dev/hda2             2.0G  914M  1.2G  45% /windows/D
shmfs                 489M     0  489M   0% /dev/shm
/dev/hda3              53G   27G   24G  53% /mnt/redhat
lethe:~/redhat # mkdir /windows/C/redhat
mkdir: cannot create directory `/windows/C/redhat': Read-only file system
lethe:~/redhat # 

I'm not going to mount it read-write yet, without Googling around to see what I'd be getting into. It's not that I care about the Windows stuff. I just don't want to lose my backup files now.

Maybe the new modules let you write safely. Maybe Sun can now get this straightened out given our new relationship with Microsoft. In any case, I don't want to go on a wild goose chase today.

I had it to be able to work on FrameMaker, for which we don't have a Linux version, only Windows and Solaris. Maybe I should try installing Solaris x86. Whenever I get out of this current mess.

Posted by Mark at 06:52 AM

June 29, 2004

ide=nodma

Right now, I'm copying about 1.8 GB of stuff from one partition to another. This kernel booted as follows:

title failsafe
   kernel (hd0,5)/vmlinuz.shipped root=/dev/hda8 ide=nodma apm=off \
   acpi=off vga=normal nosmp noapic maxcpus=0 3
   initrd (hd0,5)/initrd.shipped

The killer at this point is ide=nodma. Seems to have slowed the disk down about 1000x. Funny how there can be system bottlenecks like that even today.

Posted by Mark at 10:38 PM

Painful

Well, I mentioned earlier today that I made a little mistake with an extra /. It looks like I have at least 9 GB of stuff I wish I'd backed up, and not enough space to do it cleanly on any single partition. Maybe I can copy over Ethernet to the laptop, which seems to have plenty of space.

It looks like what went first was /usr/bin, so the system is possibly beyond the point where I can repair it faster manually than reinstalling and reconfiguring. I'm reduced to using the bridled version of JDS that sort of works on this machine.

A full backup of the partition would've been 37 GB, and I'd have had to have a good backup policy even with the equipment to back that up. Perhaps it makes sense to buy a second 80 GB disk and just copy the whole thing every week. Too late now, in any case. Besides, if I had that much space, I'd just fill it up, or do a third rm mistake...

Posted by Mark at 10:31 PM

Why doesn't rm have undo?

That makes two times in my life that I've really done something stupid with rm at home where the system's not backed up. Once I accidentally deleted a bunch of files for a project I was developing. Today I accidentally said /usr when I meant usr... as root. Hope I can find the old Red Hat 9 CDs somewhere. Who knows how much time an extra / will cost?

Posted by Mark at 06:46 AM

June 28, 2004

Google and Online Help

As you know from reading my blog, I spent some time leading a workgroup defining a common approach to online help. JavaHelp 2 offers both server-side and client-side context sensitive online help with lots of trimmings.

Trouble is, I think I quite using online help when I left Windows and got Internet connectivity. At that point, I started reading web pages.

A couple of years ago, the situation got even more titled in favor of the web with the advent of Google. With Google's advanced search capabilities, even with just "I'm Feeling Lucky," it takes me less time to look up the answer to my question in some mailing list than to search through oodles of "help" somebody threw together to explain their application before anybody used it.

In that vein, I once suggested to the group that unless you can access the online help over the web, it's probably nearly worthless. I wonder how close to true that is, statistically speaking.

Posted by Mark at 09:20 PM

June 26, 2004

You can count on Debian

Back in the late 90s, I bought a Gateway laptop. That was before we had cables at work to lock laptops to our desks. The laptop was stolen. We never found out by whom, though you had to badge to get into the building.

Anyway, since this happened at work, Sun was able to have it replaced, so on recommendation I got a Sony Vaio. The Sony laptop of course makes a good first impression, what you'd expect from Sony. It also has some flaws: the keyboard's too small; the battery only lasted about 3 months; the sound card only lasted about 4 months; you sometimes get electric shocks when you touch it in the wrong place; the surface above the processor heats up such that you can no longer rest your palms while typing after the laptop's been running for half an hour; much of the hardware remains unsupported on Linux because Sony refuses to release the specs; etc. But it sure looks hip. Almost as hip as an Apple portable.

I'd installed Java Desktop System on the Sony laptop. That went fine. Fastest Linux install around. But I couldn't figure out how to read from a little memory stick we got to take pictures with a Sony digital camcorder. So I decided to put the factory version of Windows 98 back on and try that.

I followed Sony's instructions, using the recovery CD. This of course didn't work, since nobody expected you to recover (sic) to Windows after being smart enough to partition the whole drive for Linux. ("Cannot find C:... call tech support.")

I then tried Java Desktop System. Version 1 did not target laptops. (I know this because I spoke with the guys at Sun who built it on top of SuSE enterprise edition. They were aiming first at workstations.) Java Desktop System wouldn't let me repartition, or at least I couldn't see how to do it, and when it loaded modules, Java Desktop System made itself incapable of reading from the CD it had just loaded the kernel and install program from, which is attached by PCMCIA and the BIOS knows how to read at boot time.

So I rooted around, and there were the CDs I had burned to install Debian Woody. It's true, I'm just not enough of a geek to appreciate Debian entirely. Right at the install, the Debian folk put you in charge. (For those of you who like analogies, Debian is a distro from the future where we've made the transition to participatory economics and people expect to take charge, so the Debian installer and all the rest of it seems very natural.) I had faith that when all else failed, free (as in libre) software would save me from bad engineering. I plopped the first Woody CD in the drive and rebooted.

No problem. In 30 seconds I was in cfdisk. In 2 minutes I'd partitioned the drive for Windows 95 Fat 32. Debian not only saved me, it didn't even complain that it had let me recover the capability to install a proprietary OS with a bunch of proprietary drivers whose engineers needed assistance from a bunch of free software dudes just to get their own software to function.

Can we attribute all this to GPL? Well, only after a fashion. It seems to have more to do with why the folks making Debian keep going. They love what they're doing. The engineers at Sony probably like what they're doing, but they're getting paid to do it, and many of them probably would not do it the same way if they weren't getting paid. I'd pick labors of love over labors for money every time I have a choice. Imagine a world where people work hard on things that they want to do, share the chores, and don't mind taking charge. I'm pretty sure the Debian folk already live there.

Posted by Mark at 06:51 AM

June 19, 2004

Is this for Real? (cont'd)

Ok. I'm stumped.

I downloaded the .rpm, forced it to install the beta realplay client over the working realplay client. At least it works. But it looks like only the audio got rolled into the file I generated!

When I run mplayer on the .avi version I used to generate the .rm, I get both sound an video. (mplayer absolutely chokes on the .rm, but does say the following:

Video: no video

when I run it.) What's up with RealProducer Basic?

I give up for tonight...

Posted by Mark at 11:42 PM

Is this for Real?

RealProducer 10 runs on Red Hat for me, but unless you pay almost $200, you cannot encode in anything but Real 10 format.

I created a Real 10 version of the movie burned to VCD this morning. (Watch out: 11 MB file behind that link! Still uploading as I write this.) Of course that won't play properly on my realplay 8 for on Red Hat. All I get is the audio.

So I tried downloading the beta player for Linux. I got the Helix player first -- which doesn't recognize Real 10 format! At the bottom of the page, I saw the RealPlayer 10 beta, so I got those 6+ MB of stuff, and the file wouldn't extract.

Anyway, let me know if you read this, try out the video, and it works for you.

Posted by Mark at 11:19 PM

June 14, 2004

Not Kino leakage

Fabio explained to me today that Linux caches very aggressively, so what I'm seeing is the kernel filling up the pages in memory with digital video. The content in cache doesn't go away because the kernel doesn't flush it once it's loaded, until it's displaced by another application.

I'm not sure I understand why things slow down later. Maybe it has something to do with how the kernel takes care of a pretty large number of pages in cache.

If I live to be 1000, I promise to read the kernel source before I die...

Posted by Mark at 09:22 PM

June 12, 2004

Open source online help

Roger Brinkley mentioned the other day in a meeting that we really ought to think about supporting online help for open source projects. But I realize that since I built my first working kernel, I can perhaps count the number of times I've really needed online help on Linux on the fingers of one hand, as opposed to a web browser or a man page.

But Binky is probably right for people using IDEs like Rave or NetBeans. Paul Fussell was showing off the cool help feature they've put into Rave, as sort of always-present commentary that changes as the focus moves around. The idea is that JavaHelp will go out as an open source offering... all we need to do is get Java everywhere you've got X.

Posted by Mark at 01:52 PM

Kino leakage?

My PC at home has 1 GB RAM. I'm learning to edit video using Kino as the editing tool. Pretty cool so far.

Dude, something in there leaks.

Before I start Kino, I have about 200 MB memory in active use, and about 300 MB total according to top. After loading Kino, doing very minor editing on an hour of crap I shot before I knew what I was doing, and then shutting Kino down, top says I still have about 200 MB memory in active use, but 987 MB in use. Why doesn't the kernel reclaim all that inactive memory? Are we talking serious leakage?

And why don't I get to the swap point? I've only got 1.5 MB of swap in use.

Posted by Mark at 09:44 AM

Rise of blogging

Estimates of the number of blogs out there vary. I thought Tim Bray guessed something like 2 million yesterday. A quick Google search led me to Perseus Blog Survey, where the estimate stands at 4.12 million, on the order of a blog for every 1000th human.

It struck me yesterday as I drove home after the talk that blogging answers two questions about personal websites reasonably well:


  1. How do I keep the content fresh?

  2. What lets me do that with almost no effort?

By definition, a blog answers the first question more easily than anything else out there, except perhaps randomly generated content. (And even I don't read my randomly generated content very often.)

These days, blogging software such as the stuff I'm using right now makes it very, very easy. You need to know how to read, and you need to know how to type. But that's about it.

Two improvements that together might take fast growth to explosion:

Posted by Mark at 06:45 AM

June 11, 2004

Still invisible

blog site:mcraig.org turns up no hits at all. Remind me not to go into sales ;-) I could however perhaps have a second career in bird watching or burglary...

Posted by Mark at 11:03 PM

Bloglines

Ludo's been using Juicy News Network internally at work to read weblogs. I stopped when jnn went on a rampage and slashdot.org cut me off for overscanning the news.

So I've gone to try bloglines.com. Now will I actually feel I've learned something each day from all those headlines?

Posted by Mark at 10:58 PM

June 08, 2004

No more oil

Nath called this morning to say she'd taken her car to the garage. Bricali diagnosed one problem quickly: he found no oil in the motor. He lent Nath his courtesy car, and told her something about how using cheap, no-brand motor oil might have caused this. (And I don't think Bricali had to study psychology to make that sales call.)

Coincidence: This happened once before with the same car. Coincidence: Nath's father and two brothers like cars so much, they rebuilt a late 60s Matra, including the parts, by hand in their garage. Their sister, an ESFJ, does not check the oil, even after the little red lights go off indicating the problem.

Zen and the Art of Motorcycle Maintenance comes to mind as an explanation. Sun sold so long almost directly to engineers. Now we're selling to the CIO or CEO. These guys pay someone else to check their oil for them preemptively, but they no longer want to pay very much.

On the one hand, you can understand how Leary thought of H.O.M.E.s while sitting in solitary confinement at the bottom of Folsom Prison. On the other hand, you realize how difficult we may have (not just Sun, everybody) selling software to people who don't want involvement with technology.

Posted by Mark at 09:41 AM

June 07, 2004

Virtualizing away from the disk?

In my entry entitled Learning JNDI I wonder whether we're virtualizing away from the disk. That is to say, when the disk access rate no longer holds up the rest of the server, will moving the database to memory for faster access help?

With a 64-bit server, we already support holding all data in memory cache. That memory cache remains volatile across server restarts. The disk therefore becomes a bottleneck when you write new data, and as we suggest you write that data to disk for safety.

Perhaps we could allow for duplicated, networked memory caches? We've seen it done in replicated databases. That would probably cost a lot to implement correctly.

Posted by Mark at 08:23 AM

June 03, 2004

LDAP schema registry

What would an LDAP schema registry do?

Would it follow The Schema Registry project at http://www.daasi.de/services/SchemaReg/? Is that not heavier than what we want to do?

Would it implement the MIME retreival, the search and browse interface? How would it allow versioning and change control? Would we associate schema-paks with specific versions of products?

Posted by Mark at 10:07 PM

May 31, 2004

Webcam

Apparently, webcams work by repeatedly uploading JPEG photos to a website, and having the client browser refresh repeatedly. I imagined they'd instead have a directory for peer-to-peer connection. Does instant messaging work by upload/download as well?

I guess the upload/download way of doing things makes sense, however, from a design point of view. I was imagining two peers with webcams having a videoconf conversation, but probably 99% of the webcam traffic is porn, with some lonely guy watching canned or (more expensively) live sex videos. The model for that is broadcasting, and handling all the peers would be too much. If you want high quality video, get a faster connection.

I also looked quickly at webcams, on Amazon.com for the reviews. For $40, you can buy a webcam at Amazon, one that reviewers claims works fairly well. Finally we have wireless telephones that don't require you to stand still next to wall outlet, and along come webcams. Pretty soon, you'll have to sit still in good lighting with the processor and main fans whirling in the background.

Posted by Mark at 02:06 PM

May 26, 2004

Daily average 80 hits

Movable Type must do a fair amount of traffic with the server. I imagine Webalizer inflates, rather than moderates it's average daily hits reports. But mcraig.org traffic went to 80 hits an average day when I started using Movable Type to blog.

This is up from an average of 42/day in April, when I started trying to figure out what I wanted to put up on our website. Even the spike induced by my brother in July and August of 2003 was much lower (33/day, 37/day). I haven't looked at what might be causing this.

Either that or it has nothing to do with us, and instead represents somebody hijacking our site as a source of spam. How do I read the server logs?

Posted by Mark at 09:14 AM

May 17, 2004

Ugly fonts

Perhaps you are experiencing them as well? Ever since I upgraded to Mozilla 1.6, my fonts have been hideous. But then, Palatino looks bad even in Opera on Red Hat 9.

Palatino seems to be the default recommendation for my this weblog, according to the CSS. Probably looks great on a Mac.

The fonts in my resume look disasterous as well. Could be something to fix before I try to go get another job someday.

Does anyone else out there get the feeling that this is entirely too complicated on Linux, and probably UNIXes as well? Someday I'll have to read the story of how it got like that.

Posted by Mark at 10:46 PM

Ask The Oracle

Now instead of throwing three coins or searching for a hexagram thrower at google, you can use mine. (I love pretending that someone else might read my blog someday.)

Ask The Oracle gives you another way to tell your fortune, in a fashion infinitely more inscrutable than fortune(6), because you have to understand not just the hexagrams, but also their interpretation.

So far, I've asked the Oracle once, and received the Chung Fu (also spelled as Kung Fu) hexagram in response. Maybe my question itself had no meaning. At any rate, I'm not sure what to make of the answer.

Posted by Mark at 04:21 PM

May 16, 2004

Fun with hexagrams

In playing around with the hexagrams, I've now noticed that the Wilhelm translation, mentioned at http://hexadecimal.uoregon.edu/ching/src.html does not have unique English spellings for all the hexagrams.

This breaks my script, or rather, my script breaks on that feature of the list. The Legge translation (Bantam Ed., 7th printing, picked up used in Menlo Park for $1.50) has the same feature. I suppose I'll have to distinguish with the additional text, or numbers, or something.

That aside, I'm close to having XHTML for each of the hexagrams, with the appropriate bits of the Wilhelm translation. Has anyone written a program that interprets the outcomes of its own inquiries to the oracle?

Posted by Mark at 10:22 PM

May 14, 2004

Fun with CSS

In an effort to relearn some Perl, I've started writing a CGI similar to ching(6). In itself that has nothing to do with CSS. But ching(6) in text format has the hexagrams a nice ASCII art, with each trigram labelled inline. It looks for example like:

          --------
          --------     above     Ch'ien The Creative, Heaven
          --------
          --------
          --------     below     Ch'ien The Creative, Heaven
          --------

Having redone the hexagrams in PNG format, I'd like to get the same effect... with CSS if possible. I've not been able to layout the table correctly:











Ch'ien

above

Ch'ien The Creative, Heaven

below

Ch'ien The Creative, Heaven

I don't want to resort to each line of the hexagram being it's own image, as that spaces it too much. That's what was done for the HTML version of ching(6).

So how do I do this properly in CSS, knowing that my hexagams are always 128 px wide and 94 px high?

Posted by Mark at 10:15 PM

May 08, 2004

ChorusOS and Jaluna

Sun acquired Chorus Systems in late 1997, apparently with the notion of using the Chorus microkernel as the basis for a Java operating system, but also to help Sun expand into the Network Equipment Provider market. The Chorus microkernel offers real time capability, hardware abstraction, and virtual memory. Above the configurable microkernel layer, Chorus provides support for distributed systems and different operating system personalities. In other words, you can have an incarnation such as ChorusOS 4.x, with a POSIX personality based on Free BSD, that gives you API compatibility with other UNIX systems, making it easier for Network Equipment Providers to have applications running all the way from the base station controller into the large backend systems, using the same code. Other incarnations provided PSOS APIs, for example. This has probably enabled compatibility with VxWorks as well, and of course means that you can port Linux on top of the microkernel.

At the end of the last century, the Network Equipment Provider market was expanding quickly. Sun folded ChorusOS into the Netra HA Suite for customers in this market. Netra HA Suite provides a level of API compatibility between ChorusOS and Solaris systems that lets NEPs write once, more or less, for applications running across their network infrastructure. Netra HA Suite also provides a distributed, container-based architecture for high availability across the network infrastructure. In other words, NEPs writing to Netra HA Suite can have applications fail over quickly and statefully from one hardware node to another, over the network, no matter what.

When the NEP bubble popped, and Sun was not in the market for devices, the folks who invented the Chorus microkernel went on to found Jaluna. Jaluna appears to have opened the source for the basic system, gone more heavily into Linux, and aimed at providing more functionality at the high end. An interesting historical wrinkle in all this: Tilly Bayard-Richard managed to take all the documentation we wrote in or had moved to Sun's SolBook format, a subset of DocBook, and use open source tools to rebrand and republish it without extensive modification to the content. You can see the results here at Jaluna.

Posted by Mark at 07:13 AM

May 07, 2004

Lightweight DS Install Doc?

Sun's current Directory Server inherited code primarily from Netscape Directory Server, but also took some pieces from Sun Directory Server, and from Innosoft Directory Server. As I understand it, all of these inherited code from the original UMich Directory Server. I write product documentation for Sun's Directory Server, an in particular, I've written installation documentation for the product.

For Sun's Java™ Enterprise System 2004Q2 release, we've done away with installation documentation in the Directory Server documentation set itself. Instead, everything sits in the Java Enterprise System Installation Guide. In a way, this looks like a good thing for people taking Java Enterprise System as an integrated software stack.

For those just checking out Directory Server, especially on Linux, it's crazy. One antidote might be to write a Sun DS Evaluation HOWTO.

Sun DS Evaluation HOWTO (Outline, partially stolen from the LDAP Linux HOWTO)


  1. Quick Start (willing to accept defaults to get it running ASAP)

  2. Introduction to Directory Server (LDAP, history, features, performance, scalability, non-free)

  3. Installing Directory Server (bits, components, layout)

  4. Configuring Directory Server (architecture, local/remote mgmt., perf. tuning)

  5. Running Directory Server (start/stop, console, monitoring)

  6. Directory Data (Schema, DITs, LDIF, import/export, backup/restore, replication, grouping)

  7. Directory Access (ACI/ACL, password policies, accounts, SSL, SASL)

  8. Directory Tools (ldap*, perf. tuning tools, JNDI, LDAP C API, plug-in API)

Posted by Mark at 11:35 PM

May 06, 2004

Encryption and convenience

Matt and I fiddled with gpg to authenticate and even encrypt messages. The p is for "pain." Maybe the last g is for "groin."

Part of the problem lies in that encryption cannot work if you cannot read the doc. And the doc really needs to assume you know nothing about it until you know, after which point you needn't read anything other than the man page or the code.

Of course the doc goes along as if you understood the underlying technology, which maybe you did when you read about it in a textbook, but you certainly cannot remember that now. One could get paranoid thinking about how easy it is to preven people from using encryption and authentication.

Posted by Mark at 11:31 PM

Using BlogEd

Someone once said speech originated as we had no other way to complain. Writing therefore developed not to keep track of how much we owed each other, but instead to let us refine our complaints.

James Gosling wrote BlogEd on the Mac. His README at the top of the CVS tree says, "I've only ever tried this thing out on MAC OSX, so your mileage elsewhere may vary." My mileage seems pretty good on Sun's Java Desktop System, although the HTML generated looks like it was designed by a developer, rather than a web master. It's not bad on Red Hat 9, but dragging and dropping images does not appear to work.

So why complain? I have the code on my disk, right? Yes. It looks long. I feel tired.

That said, I probably ought to read the code. If you end up having to roll your own tool, you ought to know how a few work before you start rolling.

Posted by Mark at 11:25 PM

Using Movable Type

TWiki did not work for me. BlogEd is yet another client. Besides Movable Type has these cool pastel CSS defaults.

Installing this at our current ISP with no ssh -- ssh costs more than having the site itself hosted -- worked fairly quickly, taking me in total probably 1:30. I'm not particularly fast, so sat there staring at the 403 (access forbidden) error when Apache would not run mt-check.cgi. Finally, my sleepy brain decided to recall sudo /etc/init.d/httpd start, and I looked through the doc. It turns out you can drop an .htaccess file into the directory where you installed Movable Type and Apache runs the CGIs.

$ cat .htaccess
Options +ExecCGI
AddHandler cgi-script .cgi

BTW, I downloaded the full version + libs. This gives you all the Perl modules you need to run Movable Type so you don't have to install them on the ISP's system.

I've used BlogEd at work, and also at a test site. It's not bad, and it's in Java, so theoretically, I should be able to fix the small bugs myself. The code can be had through CVS. Trouble is, I want to blog, not write a blog editor. At least not yet.

TWiki looks good for categorization and collaboration. I envisioned a family site, with multiple players. But the troubles started when I tried to get a version of Net::SMTP working so TWiki could send notifications. I even got a notification or two by email, but finally just gave up. Comment if you like. When you cannot login to your ISP, you're pretty limited.

Posted by Mark at 11:21 PM