As most may have noted i kind of "spammed" planet.debian.org today. Sorry about that. By request of Murray i tried to "upgrade" my blosxom feed to RSS 2.0... i downloaded the PyBlosxom flavour he recommended, installed it - ran it through the validator and spend quite some time in fixing it...

Well, unfortunately this pyblosxom flavour also used different path names for the blog entries. And planet.debian.org fetched the RSS before i had noticed and fixed that.

The path names are now the same as before, the date is RSS compliant, and the feed validates in two rss validators i tired.

I hope you get my old entries out of your feed readers soon... ;-)

Today i helped a friend installing Debian on a Samsung P35 notebook via ICQ.

Apparently the installation (using the LinuxTag DVD Edition of sarge) was pretty tough. First of all, three boot options were needed to actually get the installer booting. ACPI and PCMCIA needed to be disabled and the video mode needed changing, too.

After that (and adding the acpi=off parameter to the grub menu again) the system came up and the rest of the install was okay.

But gdm didn't come up, too. Seems like the "ati" driver doesn't support this new radeon yet. Vesa is now running okay but in a low resolution.

I know that the LinuxTag DVD is a bit old by now, maybe it would have been a lot smoother using the latest installer.

Still: *Hardware support in the debian installer is essential*

Especially the lockups with certain PCMCIA hardware suck (i also encountered them on a reinstall of my old Dell Inspiron 8100, ~3 years old)

The centrino wlan module probably won't be supported using packages...

I get the same result by pressing "tab" just twice.

First press is equal to pressing "?" for me, resulting in the contents of my main mail folder (including all the evolution index files etc.) and pressing tab again gives me the list of monitored mailboxes, unfortunately only indicating by "N" if there are new mails in the "new/" subdirectory, not caring about unread mails in "cur/"

Thanks [http://antti-juhani.kaijanaho.info/blog/en/stuff/mutt.html Antti-Juhani Kaijanaho] for your follow up on my blog entry.

The "nice folder overview" is pretty useless IMHO. At least with maildir i can't see any marker showing me the number of unread mails. it does only show me a single "N" for new mails. Which aren't "new" for a long time, since evolution moves them to "unread" automatically.

(BTW: pressing tab three times gives me the same result as pressing it once)

As for the "modal" thing. I wouldn't call it modal. Modal is what i call editing in vi. I would it call "single-tasked". Not too untypical for a console program.

I have less trouble with that behaviour. I know my keys to "postpone" a message, and i can just run a second mutt instance, too.

In fact what i really like about mutt is that i can use vim for editing. I can't think of any sane way to have a proper vim in mutt and have the possibility to read mail at the same time as i'm composing one.

I have to join [http://www.livejournal.com/users/wouterverhelst/33972.html Wouter Verhelst] and [http://raw.no/personal/blog/tech/2004-08-19-01-40_email_clients.html Tollef Fog Heen] on the topic of email clients.

I'm using offlineimap to synchronize my imap mailbox with a local maildir. (unfortunately, offlineimap sometimes had errors transferring emails, maybe it even lost a mail for me once, i'm not sure)

I can then use maildir-capable applications to access the mailbox at the same time which is great: as tfheen and wouter have noted, there is no perfect email client around yet.

I prefer mutt and evolution currently. Mutt is what i use on my mail server when i'm not at home. At home i use evolution and mutt.

Mutt has lacking folder support. I have it scan my important folders for new mail and when i press the change-folder button it will then show me the next folder with unread mails as suggestion. That is okay, but i'd really like to have a folder overview.

Evolution has this good folder overview. But it's lacking in other areas. With a bit of manual gconf editing i managed to get my "signify" signatures into evolution properly, but its missing a lot of scripting capabilities. For example in mutt i have identity defaults for my folders and for certain recipients. I.e. when i send an email to some @debian.org address, mutt will automatically choose my @debian.org identity.

Oh, and i can use vim to write my emails in mutt. I tried the vim bonobo embedding to use vim in evolution once, but it didn't work too great.

A long time ago i had a look at sylpheed. But for a long time sylpheed didn't have maildir support, so it was not an option. I believe that new -claws version do have maildir support i'll have to check.

Mozilla Thunderbird probably hasn't got Maildir support, but i could use it over imap probably. Still i'm as opposed to "mozilla" stuff as wouter. In my opinion, mozilla user interfaces are crap. They are hacks, ugly hacks. I do not care at all if an application looks the same running windows or linux, since i do not run windows. I prefer having an application using the same widgets as all my other applications (in my case: gtk), and i really appreciate applications to be well-integrated (in my case: in gnome)

I don't use the mozilla browser or firefox either. I use galeon, and my second choice would probably be epiphany. I still hope for someone to make gtkhtml full-fledged web-browsing capable, but that would probably require a fork (maybe you do not want a full-fledged browser to display local help files, but have different requirements)

One thing with email clients is the old "illness" called "featureitis".

I dislike applications trying to do everthing but not doing anything perfectly. I see no reason why email, calender and todo should be inside one application. tfheen noted that he expects his email client to have good filtering capabilites. I do not want these in my email client. I do Email filtering on the server. Oh, and i do not need mail fetching either.

Maybe instead of adding another new email client to this list we should think about making this more modular. Evolution already has a so-called "data server", which we could probably use to access mails as well as contacts. We'd then need probably a shell which can display the folders nicely, some email viewer component and an editor component. Eventually, each part should be exchangeable. (In fact i think evolution already does so.) - and we need a script language to automate actions between these shells. Lua could be an interesting choice, IIRC this language has been designed as a scripting interface language, easy to learn, but unlike some basic clones a more modern language. Or maybe python, albeit i find the "indentation is significant" still odd.

On the SELinux mailing list, openssh 3.9 changelog was quoted:

* Implemented the ability to pass selected environment variables between the client and the server. See "AcceptEnv" in sshd_config(5) and "SendEnv" in ssh_config(5) for details

This has potential: it could finally solve the charset issues that annoy me (i blogged about that earlier).

Unfortunately you can't just forward LC_ALL blindly. But i can deal with that in my login scripts. I just need a way to get the charset information from the client to the server.

The release managers are going to kill me.

At least if [http://packages.debian.org/modlogan modlogan] makes it into testing in time and any problem shows up.

I've been running a backport of modlogan on woody for some time, and this version has bugs. Ugly bugs. Every now and then it just eats up cpu cycles until i notice and nuke it. Well, there have been a couple of upstream releases in the meantime, including some fixes i contributed. I don't want that version in sarge.

This is why i decided to hijack modlogan on a short term: the last maintainer upload was more than a year ago, the application build process is clean (standard autotools cdbs build now, i only moved the libraries to a subdir to not clutter the normal lib path and replaced a few configuration files with better example files)

During my testing of the package i also fixed two minor things: i removed some useless but annoying debug output and fixed a segfault when run with an invalid configuration. Oh, and i added better OS and Browser detection strings.

Today, my xchat and sodipodi NMUs were installed after a 15-day grace period (i was away for a week and didn't want to have them installed before i return)

Being bored i also built xchat 2.4.0 (fixed a minor bug, too) and just uploaded them to experimental for testing. Changes seem to be minor (i.e. the version number jump is just cosmetic, probably to catch up with gnome; news include a marker line up to which has been read before, a new color scheme, dialog HIGgification and sorted tabs) so i believe this version could go into sarge unless there appear bigger problems.

So if you are a xchat user, could you test the xchat 2.4.0-0.1 packages from experimental (uploaded for sid, built on a sid chroot pbuilder)

Oh, and i was rated a "SECL--Sober Emotional Constructive Leader", "a politician" when i did that other test some time ago.

Last week was very cool, giving a workshop on "interactive art" aka "Art & computers" to some gifted youths about my age. It was really cool: although few of them had programming experience - many of them came from the art field, some with no programming experience at all - we were able to do some really cool stuff with Python, PyGame, Python-Soya, Blender and Flash.

Also i managed to find a girl who gave a short course to Rock'n'Roll together with me where up to 15 people attended. After the week we had shown them all moves we had learned ourselves yet.
I really love Rock'n'Roll although it can be really exhausing.
Now i need a girl to learn Salsa and Tango Argentino with. ;-)

The week was just great, i met a lot of good friends, found as many new friends and enjoyed it very much. Now i need to finish my project thesis, apply for going to france for half a year and then i maybe have more time for Debian again. (In case you are from France: i'd like to go to INRIA or LORIA for the [http://rewerse.net/ Rewerse] Project. Maybe we can meet then and sign keys. Intended time to go there is next year for half a year.)

Well-known [http://www.heise.de/ix/ iX Magazine] (sister magazine of c't) is doing a [http://www.heise.de/ix/linux-optimal/ reader survey] about the "optimal" linux server. (Sorry, questionnaire in german only)

The questionnaire had some really crazy questions (like which webbrowser should be installed listing mozilla, konqueror, opera but not links/w3m/lynx) and even asked wheter one wants to have openoffice and evolution installed.

The last few questions were like "the default installation is Debian Linux, which other Linux distributions would you be interested in", "do you need technical support for this other linux distribution" and "how much would you pay for technical support for other distributions"

I think i have blogged about the UTF-8 troubles before (zsh not supporting utf8, working with multiple charsets). A new idea sprung into my mind i'd like to discuss: Allowing applications to switch the charset for xterm. Gnome-terminal for example allows you to switch charsets easily (using the menubar, unfortunately not by using profiles).

Why shouldn't the application trigger this charset switch?

I'm thinking along the lines for switching the window title in xterm. Why not introduce a new "ansi-like" escape sequence that switches the charset?

I am aware that this has some problems such as applications not restoring the character set when terminated. Still being able to "echo \e]charset=iso8859-1;" or so would be cool and would allow me to customize the charset easily for different hosts i ssh onto. Restoring the charset would then be a task of your shell (for example included in the "reset" command)

But i'm way too little an expert of such things. I don't know what options of variable-passing over ssh one has that could be used to solve this more gracefully...

I'm offline for a week (ocassional access to email given) and giving a workshop on "interactive art", the biggest teaching i've yet made.

Half of the participants come from the art sector and only few have a bit of programming experience. There is no plan what we are going to do, i intend to give an overview of important things (reactivity and state models, simulation and animation techniques and such) and then try to realize some small project.

Things i really love are some of the effectv things (chameleon is great!) - i'd would be so damn cool if we were able to do something like that. Maybe a bit less technical but more art. I'm also thinking about using blender and/or python-soya or pygame. (But i have never written a python program)

If you have some idea, something i should not miss, please send me an email (we won't have a big network connection, just a Modem for ~120 people) i'd really appreciate suggestions. Also if you could point me to some good introductory papers or examples i'd be very happy.

I also do see these ssh scans on a couple of boxes. I don't think this is a distributed password cracking attempt (test is a bad choice, and even 100*8 tries is nothing)

I think this is just some new script kiddie tool scanning for known weak passwords. I've seen the usernames "user", "test" and "guest". I bet the passwords tried are similar...

Maybe Debian should disallow a certain set of _really_ bad passwords by default. I wouldn'd suggest running full cracklib, but disallowing "user", "guest", "test", "password" and "12345678" could at least discourage such scans (and prevent people from using such stupid passwords).

I was just informed by edd that my post about my intended NMUs causes crashes with a couple of RSS readers.

A bit of investigation and running rss validators revealed the following:

My RSS feed was okay. The bug is triggered by the following coincidences:

* My blog posting title contained an ampersand, correctly encoded as ANDamp;. * PlanetDebian loaded the blog entry, decoded the ampersand, but did not reencode it again in their rss feed * the used rss readers crashed upon the ampersand without a proper entity following. (they should be more fault-tolerant, i think an entity name must never start with whitespace, so the parser should have been able to recover gracefully)

Well, until this is fixed, please no not use ampersands in your blog posting titles. I don't know about the bodies, i decided to avoid them there, too, for now.

I just uploaded three NMU to 5-days delayed. Please don't kill me. ;-)

Xchat gets the "don't lose server list on disk full" patch back with a small bugfix (a variable was not initialized causing the server list to be never saved), two upstream fixes and a couple of minor things such as the "irc" virtual package. (which should be irc-client IMHO, but =apt-cache showpkg irc=)

sodipodi had a new upstream version in February with a couple of bug fixes as well as a new tool many might know already from inkscape. I also made sodipodi register itself for the appropriate mime types.

w3c-dtd-xhtml is third, actually a package which has been sitting on my hard drive since May. I wanted offline xhtml DTD verification with xsltproc and xmllint, but the package did not include an xml catalog. Well, i hacked together one that works okay. I was unable to find a howto or a tool to generate this catalog, so i just tried around a bit. I guess the catalog might contain some errors, but it works for me. Also debian has announced a long time ago that we have "integrated xml support", and xhtml missing from that is really bad, we do need some xhtml support for sarge.

If you have more knowledge about xml catalogs, please make a proper catalog and upload. Mine went to 5-days, so either the maintainer or someone with catalog skills should be able to upload faster. ;-)

Yes i know that uploading something you know you didn't do properly is bad, but first of all it worked just fine for me for months of occasional use, secondly i think this is important, and third i uploaded to delayed, there is time for someone else to do it correctly.

Happy hacking!

And please, review all of your bugs and work on these you want to see fixed in sarge. We want high quality packages in sarge...

Scott, i just don't want eds to load all the libraries. If i would not need the library otherwise this memory _would_ be free.

But it seems its not the libaries: Other gnome apps with even more libaries need way less libraries than eds.

Gaim for example does NOT show up with 80 MB virtual size, but 25 MB.

Name: evolution-data- Name: gaim VmSize: 81960 kB VmSize: 25340 kB VmLck: 0 kB VmLck: 0 kB VmRSS: 7372 kB VmRSS: 14720 kB VmData: 68856 kB VmData: 4796 kB VmStk: 20 kB VmStk: 96 kB VmExe: 188 kB VmExe: 828 kB VmLib: 10104 kB VmLib: 15336 kB

I have to admit, that gaim apparently loads a lot more libraries.

Don't tell me, the VmData are shared libraries. That is the heap.

The shared libraries are not even resident in memory, even when i'm running evolution they seem to be unneeded...

[Scott James Remnant http://www.livejournal.com/users/keybuk/13606.html] accused me of not knowing about memory usage.

I do know about shared libraries and shared memory. Still, loading libraries i wouldn't need to keep in memory otherwise does eat up memory. If you look at the ld output for evolution data server (note that i'm not talking about evolution itself!)

        libedata-book.so.1 => /usr/lib/libedata-book.so.1 (0x4002d000)
        libebook.so.8 => /usr/lib/libebook.so.8 (0x40047000)
        libedata-cal.so.5 => /usr/lib/libedata-cal.so.5 (0x40075000)
        libegroupwise.so.3 => /usr/lib/libegroupwise.so.3 (0x4009a000)
        libsoup-2.2.so.3 => /usr/lib/libsoup-2.2.so.3 (0x400ac000)
        libecal.so.6 => /usr/lib/libecal.so.6 (0x400d3000)
        libedataserver.so.3 => /usr/lib/libedataserver.so.3 (0x40172000)
        libldap.so.2 => /usr/lib/libldap.so.2 (0x4023b000)
        liblber.so.2 => /usr/lib/liblber.so.2 (0x40270000)
        libresolv.so.2 => /lib/tls/libresolv.so.2 (0x4027d000)
        libgnome-2.so.0 => /usr/lib/libgnome-2.so.0 (0x4028f000)
        libesd.so.0 => /usr/lib/libesd.so.0 (0x402a5000)
        libaudiofile.so.0 => /usr/lib/libaudiofile.so.0 (0x402ae000)
        libgnomevfs-2.so.0 => /usr/lib/libgnomevfs-2.so.0 (0x402d2000)
        libgconf-2.so.4 => /usr/lib/libgconf-2.so.4 (0x40333000)
        libbonobo-2.so.0 => /usr/lib/libbonobo-2.so.0 (0x4035d000)
        libxml2.so.2 => /usr/lib/libxml2.so.2 (0x403b8000)
        libm.so.6 => /lib/tls/libm.so.6 (0x404b3000)
        libgnutls.so.10 => /usr/lib/libgnutls.so.10 (0x404d6000)
        libtasn1.so.2 => /usr/lib/libtasn1.so.2 (0x40536000)
        libgcrypt.so.7 => /usr/lib/libgcrypt.so.7 (0x40547000)
        libnsl.so.1 => /lib/tls/libnsl.so.1 (0x40593000)
        libgpg-error.so.0 => /usr/lib/libgpg-error.so.0 (0x405a8000)
        libz.so.1 => /usr/lib/libz.so.1 (0x405ac000)
        librt.so.1 => /lib/tls/librt.so.1 (0x405bd000)
        libbonobo-activation.so.4 => /usr/lib/libbonobo-activation.so.4 (0x405c3000)
        libORBitCosNaming-2.so.0 => /usr/lib/libORBitCosNaming-2.so.0 (0x405d9000)
        libORBit-2.so.0 => /usr/lib/libORBit-2.so.0 (0x405de000)
        libpopt.so.0 => /lib/libpopt.so.0 (0x4062d000)
        libgobject-2.0.so.0 => /usr/lib/libgobject-2.0.so.0 (0x40635000)
        libgmodule-2.0.so.0 => /usr/lib/libgmodule-2.0.so.0 (0x4066f000)
        libdl.so.2 => /lib/tls/libdl.so.2 (0x40673000)
        libgthread-2.0.so.0 => /usr/lib/libgthread-2.0.so.0 (0x40677000)
        libpthread.so.0 => /lib/tls/libpthread.so.0 (0x4067c000)
        libglib-2.0.so.0 => /usr/lib/libglib-2.0.so.0 (0x4068b000)
        libc.so.6 => /lib/tls/libc.so.6 (0x40709000)
        libcrypt.so.1 => /lib/tls/libcrypt.so.1 (0x40845000)
        libsasl2.so.2 => /usr/lib/libsasl2.so.2 (0x40871000)
        libgnutls.so.11 => /usr/lib/libgnutls.so.11 (0x40886000)
        libasound.so.2 => /usr/lib/libasound.so.2 (0x408ec000)
        /lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0x40000000)
        libgcrypt.so.11 => /usr/lib/libgcrypt.so.11 (0x4099c000)

Is there a single library the data server does NOT link on my system?

I guess if i hadn't got 512 MB of RAM i wouldn't be able to run all my beloved GNOME2 stuff, galeon, xchat and the omnipresent evolution data servers...

And program startup times do depend on the amount of libraries they need to load (yes i know that evolution is starting a lot faster because the data server has already loaded all the libraries into memory)

I've been working on an MMU-less embedded ARM system for the last 10 months. Since then i've been a lot less confident in the code quality - especially wrt to memory management - of open source software. (Of course there are many good apps around, too)

I'd love to see studies how much speed difference a more conservative memory use can make on systems that are low on memory (which probably applies to desktop machines with <= 256 MB already nowadays)