Older blog entries for yeupou (starting at number 74)

Exim in Debian

One thing that I usually enjoy with Debian is the fact that many software comes the way their upstream author wanted it. KDE is a good example. I must say that, in this regard, I'm quite annoyed with exim 4. While with exim from http://www.exim.org you have one clean flat configuration file, with debian, now (unlike with exim 3.x), you have to deal with tons of small file or one big unreadable file. If you choose to use the unreadable big file, you end up forcing settings at the begin of the file because you do not now how to put them otherwise (and when you are already familiar with exim4 and have a system running, you may not be very interested in learning how this system work). And you have to careful select where you'll put your personal routers and transports between heavy blocks of hashes. Once it is done, your configuration file got migrated, stripped of the comments in /var/lib/exim4 -- a very common place for configuration files. This whole thing seems overcomplicated to me. I always preferred exim over sendmail and postfix because it was damn so easy to setup in an efficient way. But as it is shipped with Debian, it is no longer the case in my opinion; I'm more efficient at recompiling the package directly from upstream, that's not a good sign.

One other funny thing is the fact that instead of using the mail or exim account, exim 4 use a user called Debian-exim. The "Debian" is quite interesting: I am running debian, do I need that to be recalled everything? The default output of ps is now very informative (user = Debian-[squish]), I'll be very happy the day apache will be called Debian-[squish] too. The capitalization of the first letter is nice too: it is the only system account that have it.

I am sure there are many good reasons behind this, even for the strange account naming policy. But I am truly unsatisfied. I have the feeling of running a package that was imported from another distro, an alien package, that obey to a different set of rules than the others.

Debian Sarge

One thing I value most in Debian is the distinction between stable, testing and unstable. I did not agree with people saying stable should be released for often so it fits for desktop usage. Desktop usage can handle a few bugs and frequent changes but not servers: if people want to use the latest stuff, going to test (or even unstable) should not be a problem. Since two years, I use testing on my own desktops machines/laptop, I never suffered serious package upgrade problems. But now we reach a point where stable is no longer suitable even for servers. If you run a mail server, you'll probably want to have spamassassin, maybe mhonarc and mailman: these software in stable are so outdated that you are likely to start with testing. If you run a webserver, you'll maybe want to test apache 2.0 ; let's go for testing too. If you run a subversion server, you have to go for testing...

Comparing arch to BitKeeper

A while since BitKeeper was selected for Linux development, there are proposals to switch to a free software solution like arch.

While Linus Torvalds starts arguing that CVS is bad, some others like Jeff Garzik give us the good ol' "if both BK and <open source tool> were completely equal in terms of function, I'd use the open source tool" argument. This last stand is quite funny since it is typical self-blocker logic. Considering that you need a certain amount of users to be able to enhance your software (especially for a kernel, unless you are wealthy enough to buy yourself all kind of devices only for testing/writing modules purpose), if users follow this logic, your software will never get a chance to get improved, because they will never start using your thing. If Linux is nowadays widely used, it is specifically because many persons refused such logic; and in many cases, it continues to grow because many persons are still refusing such logic.

Apart from that, Linus Torvalds does not show himself like a smart and sensible man when he seems unable to say anything else than like "shut up" and complain about someone would be "whining" while he plainly recognize feeling insulted when someone dares questioning decisions he made ("then complaining when people decide to use the best tool available is fricking impolite"). Not to mention the fact he talks like he was paid to do washing-powder advertisements ("the best tool available") obviously without clear knowledge of the situation ("I looked at it [arch] before starting BK"... three years ago). Truly lamentable. Not to mention the fact that he confuses proprietary and commercial software (on purpose maybe) so he can say that BK license (he does not use the word license, interestingly enough) is no big deal since commercial is not a problem ("Sure, BK is commercial, but dammit, so is that 2GHz dual-G5 too and that Shuttle box in my corner"); please, dont tell me this guy worldwide known for his work on a kernel which is nowadays largely commercialized is not aware that his damn project is under a commercial license too (since it's commercialized!), so being commercial cannot be the issue with BK licensing. And how ignorant looks this guy that compares someone criticizing his choice to use BK (an attitude he calls "whining", out of neutrality) to someone who would give away his freedom by forbidding it to use BK: in this perspective, completely off-topic (did someone ever proposed to disallow him to use BK?), he invents what Voltaire would say ("I may disagree with your choice of license, but I shall defend to the death your right to choose it") without even realizing that the original Voltaire's sentence what about freedom of speech, so the freedom of criticizing, the freedom that Linus, full of respect for it, call "whining".

Kernel-traffic is fun.

AlanHorkan:

"For one thing the Inkscape developers are no less independant than the Sodipodi developers."

Someone working at the OSDL, publishing an article about his fork, clearly is less independant than the other and clearly got a more important advertising power. That's a fact, isn't it? The question is more *was this advertising power used to screw sodipodi or not*.

" It is unfair to claim the Inkscape developers have screwed Sodipodi. They have forked Sodipodi exactly as GNU General Public License (GPL) entitles them to do"

Nobody said they have no right to fork.

"Inkscape is not in competition with Sodipodi"

This is not the point of view of the Sodipodi authors. I was not convinced either by what I read on Inkscape website. But I am not so interested about this conflict; so you may well be right.

"It is not surprising that any project bemoans the existance of a fork because if your project has been forked it makes you look really bad. [...] The Inkscape developers are very much aware that they would not be where they are today without having the solid base of Sodipodi to build on. I can understand Lauris Kaplinski's disappointment at Sodipodi being forked but you shouldn't hold that against Inkscape. "

1. I have nothing against the concept of forking. 2. I hold nothing against Inkscape.

Inkscape or the right to fork was not really the issue I was wondering about. I'm not sure it is necessary to continue a discussion focused on that here (or an article should be published) and I am frankly not really interested in the specifics.

I'm just wondering about the risk that "the ones with more advertizing power [can] screw independent [free software] developers" -- but is probably true in many others areas, so maybe it is not even software this issue is about.

Reading this message was a quite unpleasant experience, as it leaded me to take into consideration an issue I have no solution for, the fact that, even with free software, "the ones with more advertizing power always screw independent developers". The rest of the thread is a bit regovirating, but only a bit.

It's always funny to browse with google. Frequently you find what you were looking for; sometimes even more :)

Today, looking for something related to arch, I've just found out that some persons (mostly related to GNOME/Hurd, whatever, using Savannah) are not really sure if I am a "gros chieur" or a "petit con". A guy named "dodji" would go for the first one. Another, called "manuel" would pick the second option.

The dodji guy is likely to be sdodji, member of the project mlview. I do not know exactly who he is, what he wants. The only thing I can tell is the fact that I several handled his support requests at savannah.

This manuel is likely to be manuel menal, a guy I'v met on IRC a several times, several years ago. But this guy "knows me well" ("je le connais bien" he said). I do not now exactly what is his problem. Apparently it is related to Savannah but it looks like he sent private messages, while the guy he was writing to was replying publicly, so there's probably a part (that could be guessed, though) missing.

What is interesting, in my opinion, in all of this, is the fact that these guys feel entitled to insult me not because of things I would have said that are not ok in their opinion (that would not have been a surprise to me; I usually speak frankly, I do not expect to avoid harsh statements in return), but because what they apparently think they know about my involvement in Savannah.

"and there are a hell of a lot of people in the world like you, that I have seen
there are hell of a lot of people in the world like you" were singing the Newtown Neurotics. Never ever found a community immune to this.

Update: Intrigued, I found even more funny involving mmenal and Savannah. Apparently a guy was saying a project was refused at Savannah. Here comes manuel cclaiming that to get an account accepted, one just have to ask him, because he is very familiar with people at Savannah (just like if Savannah at that time had some kind of registration approval process that applied only to one part of the registrations). So in 2002, he was already such a pretentious and plainly off-topic guy. Marvellous.

What a bunch of assholes exists in France, frankly. One other guy went as far as going on irc with my pseudo (+ an underscore) saying obvious crap. This is free software in France. Well; hopefully, this is a part of free software.

Bah, I don't like humans anyway. Apart from my girlfriend, my cat and my family, I dont care about the world. Why should I?

jpick : your product seems interesting. But I cannot take a look at the ads because an annoying software requirement.

ncm, should we understand from your last post that nowadays amd64 are as well supported as pentium4 by common operating systems; and that amd64 are also as efficient for common use? If so, that's interesting.

rbultje, as long as you do not intend to declare war to make unstable part of the world way more unstable, to remove dictators that your own country supported for more than ten years without any clear evidence that you can make the situation better, I'm perfectly fine with your proposal.

The issue with Bush is not that he said he want to protect his country. The problem is how he intended to do so and what he did claiming he was doing so. To take Iraq as example, it is easily to point out that many persons warned USA the mess it would create to remove Saddam Hussein with a war -- but USA gone to war, with false statements as legitimacy. And now USA soldiers get killed every day (and the fight is still going on in Iraq, not yet moved to occidental countries). And now islamists seems to have way larger audience in Iraq than they did previously under Saddam Hussein (remember he was supported for 20 years because he was not bolchevik neither islamist, despite the fact he was already know as a criminal dictator)

So let's go, take action against islamists. But do not add a bunch of troubles ontop of the existing troubles.

ppol, in my opinion, it is a real regression when developer spend their time to choose the colors that should be used in their GUI. Modern graphical toolkits handle themes. So why not letting the themed GUI dealing with colors ?

I've read an article on netcraft about Microsoft recommending usage of passphrases instead of passwords.

The point being made is that "passwords are ridiculously easy to guess or crack", the only solution being increasing the lenght of the pass strings.

Isnt it obvious that doing so means increasing severely the pain it is for users to remembers password -- so increasing the risk of having password and stuff like that store in clear-text somewhere. While I'm definitely not a security expert (not a computing expert anyway), it seems more sensible to me to incitate to use system with key pairs authentication. On all the servers I am admin of that run sshd, only passphrase with keys permits to gain access. I think it is the way to go. That said, if many servers follow this approach, it will be necessary to disallow ssh keys pairs without passphrase (because crackers will likely being cracking users home account to steal private keys first - and that would probably be even easier than to crack a sshd production server well monitored that allow password ssh access).