Resuming my tech-notes blogging…

I’ve recently bought three Studio Ghibli’s movies. I don’t have a TV-set (yes, lenina, I’m still not watching any TV), so I’m watching DVDs on my laptop. DVDs I got are interlaced, when watching them on a computer, you can see small horizontal stripes along the sides of moving objects.

On a TV-set, this is not a problem, actually, interlacing makes animation on TV-set smoother. But on a computer, all you get from interlacing is stripes.

MPlayer comes with few deinterlacing plugins, from which yadif and kerndeint seem to work best (but I’m open to other suggestions). However, my video output is GL:

maciej@clover ~ $ cat .mplayer/config
vo = gl
autosync = 20
monitorpixelaspect = 1
af = volnorm=2:0.6
ao = pulse

The problem with those deinterlace filters is that they don’t cooperate with the gl video output. Fortunately, there’s an easy fix for that: append scale filter.

mplayer -vobsubid 1 -vf kerndeint,scale dvd://5

Replace 5 with your title number. The vobsubid option tells MPlayer to display subtitles. I like to watch Japanese movies with the original soundtrack and English subtitles. Usually, English dubbing reveals too much about the characters and lacks the original expression.

It’s a typical example: the paper is published, describing a new algorithm for data analysis. Mathematical background is described in the paper, roughly. A piece of software that implements it, is written and available for download from a web-site. You visit the web site, download it and run it. You get unexpected results. You wonder what’s happening. You go back to the site and look for the source code ― and it’s not there.

I’ve recently visited and tested two pieces of software doing basically the same thing: predicting missing genotypes. There is no source code for any of those two, and fastPHASE additionally needs you to register and accept an academic license to use it, introducing an annoying delay in obtaining the program.

By the way, why are all those scientific program names written in UPPERCASE? Because it creates an impression of IMPORTANCE? Just a side note.

Scientists work for the sake of humanity (I hope), striving to make our world a better place. Right? So why don’t they make the source code available?

Not releasing source code of scientific software is a Bad Thing, because it harms research in the field and is antisocial. The ones that lose, is the closed-source project itself, other projects in the field, and subsequently, everyone who could have benefit from the research. The only one who can possibly benefit from it, is only the author, but I highly doubt that they ever do.

Keeping the source code secret is a typical practice for corporations, who seek to profit from selling the binaries. I don’t know what business model can be built on restricted source code access in science, but I don’t think they’re every going to make any money on that.

What could be other reasons not to release the source code? Remaining the sole author, keeping all the credit? Keeping complete control? Hoping to sell license to business clients?

The main effect of making the source code unavailable is that the program internals cannot be inspected and analyzed. It’s only a binary that is available; people can obtain it and run it, without being able to modify it.

All the general arguments pro open-source software apply to the scientific software. Obstructing the software has several negative results.

• Fewer people use the program.
• None of the users can adapt or fix the program.
• Other developers cannot learn from the program, or base new work on it.

I think that should be enough, but I would like to add two points that apply specifically to scientific software.

Loss of credibility

In scientific research, they key point is to prove and verify the results. With closed source, other scientists can only run the software and examine the output, without being able to check if the program really does what the paper describes. Being unable to do that, the rest of the world has to believe the authors. Do they have something to hide?

I don’t think scientists would actually question a paper as a whole because of the source code unavailability, but it certainly makes raises some concerns about its quality.

It’s antisocial

Scientific research is usually funded from government grants, which in turn come from tax payers. Scientists are not corporations who fund themselves. It’s the society, it’s the other people who effectively pay for the research (through various funding organizations), and I believe it’s a moral obligation to, if they share their research results, share them fully.

By not releasing the source code, they only make an impression of publishing their work. They can get away with that, because many people will think that, if they can download the program and run it, it’s “available”. But it’s not!

Please, dear scientists, do what guys from projects such as GNU Octave, or R project do: share your source code. Everybody will benefit from it, including your projects and yourselves.

Since I’m teleworking, I use VoIP a lot. My usual equipment is the Creative HS300 headset, which has a cable, so I’m tied to my desk while talking. I would rather walk around, saving the sitting for later.

I recently bought a nice bluetooth headset. I got it to work partially under Linux. Initially, it works fine, but it starts to be intermittent after some time, going on and off about every two seconds. That’s probably a problem with the btsco driver. If the authors fix it, it will be great.

To satisfy my curiosity, I tried it under Windows as well. I expected it to work like a charm, but it didn’t. I installed the driver. Reboot. The bluetooth devices management application has frozen. Reboot. I got it to work second time.

The problem is, that Windows can’t manage to switch the headset between stand-by (low battery usage) and on-line (high battery usage). I have to switch it manually between stand-by and on-line. And here’s how it works (or doesn’t) with Skype:

• As the headset manual states, it should be kept in stand-by mode, which means that the headset keeps in touch with the computer, but doesn’t transmit or play any sound. It can stay in this mode for quite a long time.
• When the headset is in stand-by and I answer or make a call, I hear nothing, the headset wouldn’t go on-line by itself.
• I switch the headset ot on-line manually (during connection) but the sound wouldn’t appear. I still hear nothing.
• I have to hang up, switch the headset off, switch it back on, then make a call (or answer) again. Wow, now there’s sound!
• I finish talking and hang up. But the headset wouldn’t switch to stand-by mode by itself. It says in on-line and plays some clicks and cracks every now and then, irritating. And on-line is consuming a lot of battery, shortening the battery lifetime from 120 to about 6 hours.
• So to save the battery, I switch the headset manually to stand-by.
• …but there’s another call, I answer it and… there’s no sound!

With Linux, it is very promising to know that the btsco driver is able to automatially switch the headset between stand-by and on-line, and use the full 120-hour battery life, contrary to Windows’ 6 hours of constant clicking and cracking on-line. It’s only the matter of the intermittent sound that renders it unusable. And it must be some bug in the driver, which, unlike the Windows problem, can be probably found and fixed.

Starting from my favorite Gentoo site, I came across this site, where I found this:

“The Gentoo install was like finding a forest, cutting down the trees, breaking a leg, getting out of the hospital, coming back, taking out the stumps and clearing the land of rocks, planting wheat, building a mill, making flower, making dough, then building an oven in which to burn it to a crisp, stealing a cow, milking it, churning the butter, picking berries, making jam. YUMMY! Breakfast is ready, but you don’t have time to eat because it’s time to update everything, including GCC, which also means a complete recompilation of everything.”

ROTFL! You’ve made a good point, to a degree. But do you know what has pushed me out of Slackware?

The default system was kinda empty and I had to install a thousand libraries and compile them with no automation, and that drove me crazy. I tried SUSE, but I hate rpms and I found it unbearable. For example, looking for right rpms is very annoying, when you can’t find them for your distro, then you try some other version’s rpms, and they sometimes work, sometimes not. Suddenly it doesn’t play MP3s, or DiVX, you need to look for howtos on how to make it play those multimedia formats, then there are 15 of them, of which 13 are outdated.

Gentoo has excellent documentation and howtos. You won’t find 15 howtos on one topic. You will find only one, written well.

With Gentoo, I have a huge repository of build scripts with all the fancy stuff I like to use. All the license-related issues are left to me. If a fetch-restriction is on (no automatic downloads of Sun’s Java, for instance), I have to download the source file manually, and portage installs it for me.

I have all the recent applications, like Picasa or Google Earth in the distribution! And I have them quick.

But… um… yes, I needed to upgrade GCC once. However, it wasn’t more hassle than installing a new Slackware with new GCC, which forced me to recompile all my hand-compiled stuff again. With no automation.

I’m still using Wengophone Classic, which works well, it just has some annoying issues.  I would expect the developers to eventually fix them, but the whole development powers are invested in the Wengophone NG (a.k.a. 2.0). So those little things will keep bothering me until the new version of Wengophone is usable, which doesn’t seem to happen soon. They publish the betas, but I neither could run the binary version nor compile the source code.

The main three problems are not that big, but since they happen very often, they get more and more annoying.

• When placing a land-line call, the application freezes for about 5 to 10 seconds when the remote party picks up the phone. The processor gets at 100% use and the application doesn’t respond. After this period, I hear a time-squeezed helloing from the other side (“Hello? Hello?”) and if the other party was patient enough to wait 10 seconds without hearing any answer, I can start the conversation. Otherwise, I have to call again and again, until Wengophone doesn’t freeze on connect.
• HTTP-tunnel doesn’t work. Well, the tunnel itself does, but the application freezes after few minutes. The processor gets at 100% use and Wengophone stops responding. This time it doesn’t pass by itself. I have to kill the application.
• The ALSA sound is choppy. I’ve already written about it. I have to compile it with the OSS support.

I don’t know how hard it is to fix those issues, but since the Classic’s code has been thrown away, I wouldn’t expect the developers to drop fixing the unstable NG just to invest their time in a trash.

But wait, why trash? It’s a working application, isn’t it? It’s definitely more working than Wengophone NG. A lot of time was invested in this code and many bugs found and resolved. And now they’ve thrown it out.

Joel Spolsky wrote (in 2000):

Netscape 6.0 is finally going into its first public beta. There never was a version 5.0. The last major release, version 4.0, was released almost three years ago. Three years is an awfully long time in the Internet world. During this time, Netscape sat by, helplessly, as their market share plummeted.

It’s a bit smarmy of me to criticize them for waiting so long between releases. They didn’t do it on purpose, now, did they?

Well, yes. They did. They did it by making the single worst strategic mistake that any software company can make:

They decided to rewrite the code from scratch.

Wengophone’s case looks very alike. As a user, I’m not getting any updates of the application I use. The new version of Wengophone has still a far way to go. I think it would be a better idea to gradually replace one piece of Wengophone by another, constantly having a usable application.

Since upgrade to Ubuntu 6.06 (Dapper Drake) on my Amilo Pro V2000D, the sound doesn’t block anymore. It’s nice that all the applications can use the sound at the same time, even after switching the user — the other session is able to use the sound as well.

There is one thing that got worse after the upgrade: the sound skips sometimes. While listening to the musisc, it sometimes suddenly skips few seconds of the music. This is annoying, because then the music misses the beat and it’s destroying the pleasure of listening.

I noticed that it always (I think…) happens in the same moment: about 3 minutes 40 seconds of the song.

UPDATE: I think it’s more like 3:38. And it’s not ALSA, it Rhythmbox 0.9.5 that does that.

A friend of mine just released a converter of .dbf files to PostgreSQL, under GPL license. He writes:

This document provides a step by step instruction how to convert DBase II, DBase III+, DBase IV and (to some extent) FoxProV2 files to PostgreSQL. This solution is based on excellent Simple DBase SQL by Michael Lecuyer.

About four years ago I had to fiddle with Vim to make it edit an UTF-8 encoded file, while my terminal (and system) used ISO-8859-2.

I use UTF-8 as the default encoding today. How to edit with Vim an ISO-8859-2 file when my system uses UTF-8?

The first thought was to make it the same way as before, just swapping the parameter values: set encoding to ISO-8859-2 and termencoding to UTF-8. Unfortunately, it doesn’t work!

After some RTFM’ing, I learned about the fileencoding parameter, but setting it to ISO-8859-2 didn’t fix the problem. This was to be solved in a different way.

Vim will automatically guess the file encoding, as long as the encoding is listed in the fileencodings parameter. I use only UTF-8 AND ISO-8859-2 (alias LATIN2) encodings, so adding following line to the ~/.vimrc file fixed the problem:

set fileencodings=utf-8,latin2

Current encryption HOWTOs tell about encrypting the whole /home partition, but how to enrypt only your home directory with your own password?

My friend has written a HOWTO: Encrypted home on multiuser system.

I’ve created an ebuild for CLIPS, the tool for building expert systems. After submitting the ebuild to the Gentoo Bugzilla, I was told it was a duplicate and I should’ve searched. Well, I did, but I only found Eclipse related bugs and couldn’t filter them out to see this sole CLIPS bug.

The previously submitted CLIPS (for version 6.21) bug is written better than mine. However, it won’t compile because they released the new version with the same file names, so it’s impossible to emerge it. I updated this ebuild and if you want to install CLIPS on Gentoo, it’s available.

If you want to install it on your Gentoo, read on how to install 3rd party ebuilds, then go to the bug 49964 on Gentoo Bugzilla and download the ebuild, two patch files and the makefile. Then you need to add one line to your /etc/portage/packages.keywords file:

echo ’sci-misc/clips ~x86′ \
>> /etc/portage/packages.keywords

The .ebuild file should go to the sci-misc/clips directory, and three other files into sci-misc/clips/files. After you’ve put the files, you need to make a digest:

ebuild clips-6.24.ebuild digest

This will download the necessary files and create a digest. Now you’re ready to emerge CLIPS.

emerge clips

This command will give you two CLIPS executables: “clips” which launches the command-line interface and “xclips”, an X-Window application.