On Sunday 03 April 2005 14:39, Patrick Useldinger wrote: Hi. I doubt that slackware will die. I think there is also binary package, like bsd have. Disk space is not the problem, as compiling software take more place than using binary. And since almost all distro provides a easy way to install a compiler, not having it is not a so good protection. However, having the choice of features is interesting. If someone do not want to have ipv6 for some reason, he can disable, and it may stop some problem, either security or stability. Or if he really need to have it, he can enable it, while it would not be as easy on a binary package distribution when the maintener disable it. Not me :) I used it on a chroot, just to try. It was nice, but the lack of granularity ( ie no separation between library and header ), and my slow harddrive stopped me from keeping it up to date. However, with a decent computer, I think it may be a nice distro, something different and interesting. I do no like webforum, so I guess I would not have been able to immerge myself in the community, which is one of my main reason of using my current distro. Because you can choose the precise version of the component you use for your server. Right now, if you are using a binary package distro, like Mandrakelinux, Debian, or Fedora, you need to use the version of the package that the vendor choosed for you. For example, no way to use samba3 on woody, except by backporting. And backport is not a perfect solution when you start to mix them. Using emerge somehow ease the backporting. You can have stable software as a fondation, and you can choose to use more recent versions of some other servers, even if they are not labeled as production ready by the gentoo community ( ie unmasked ). A gentoo dev explained to me that unless debian and other distros, software are not migrated from the unstable branch to the "testing" or "stable" branch by a automated script, but people first need to do a report to say "ok, this software works here, I use this and this as flags". Once enough peoples have validated the version, it is "unmasked" by the developer, ie labeled as ready for production ( masked packages are like the unstable/developement version of others distros ). However, nothing prevent you to install a masked package on your gentoo, even if no other version are masked ( and if it works, to write a quick report, and even if it do not work ). This way, you can test the software one by one, and not the distro as a whole. Having a source based distro also mean you can have the package as soon as the source are avaliable ( and that you finish to compile ), that may be interesting in order to gain some time when testing a product or a server ( and it may be also easier to do than compiling by yourself, since you can always forget a step ). One thing that is not great is the lack of security backport. A security update is usually a update to the version where the Changelog say "security update" and no patching is done on older versions. I do not think it is good for a production server to have version updates ( like using debian testing on a critical server ). Again, this may have changed, I didn't look. -- Mickaël Scherer On the importance to respond to proposal email : http://www.nntp.perl.org/group/perl.bootstrap/1127

Hi, On Sun, 2005-04-03 at 14:39 +0200, Patrick Useldinger wrote: you mean hardware architectures I presume Debian has this too, although Gentoo gets the newest software more quickly then Debian (unless you use the right unofficial apt repositories I guess). Nobody knows :) When you wrote your first Makefile and understood how it works, you might have had the thought: "wouldn't it be cool if I could compile my whole distribution like this". I think Gentoo is the result of someone going ahead and doing it. Not me since I think the whole idea is quite ridiculous. It might be useful for example if you want to compile apache with stack-smashing protection. But then you can do that with an RPM or DEB based distribution too, it's not difficult. That's the only positive argument I could come up with :| (Rant follows) Some people like the sense of achievement that compiling /bin/true with "-O3 -leet -superfast!1" gives. But the speed benefits that can be obtained with such options are mostly in the 0-5% range, which is below the limit of perception of ordinary humans. Another fact that is mostly ignored in speed-related discussions is that for desktop use, the processor is not the bottleneck, but rather the IO system. I suspect most perceived speed benefits are due to the placebo effect. By the way, Gentoo users are quite unpopular with software maintainers. This is because they are quite vocal, they often use dodgy/overclocked hardware and experimental gcc releases, etc. Then they complain loudly when software fails on their riced-up systems. Of course these generalizations don't apply to all Gentoo users - but a few black sheep quickly give a bad reputation to the whole crowd. Obligatory related link: http://funroll-loops.org/ Cheers, Christian Glodt