Bruce Perens: the Linux colonel talks (vnunet)

Posted Sep 9, 2004 7:32 UTC (Thu) by gvy (guest, #11981) [Link]

> but when it was introduced, Fedora was

which one? ;-)

Posted Sep 9, 2004 7:54 UTC (Thu) by bojan (subscriber, #14302) [Link]

Is "sid" actually "released", in the sense Fedora is? My understanding is that sid is "always in flux", which is not the case for Fedora. With Fedora, you get milestone releases and then a set of updates for that particular release. Then, later on, Fedora Legacy (this is pure volunteer bit) takes over and makes sure security issues are fixed for the next year and a half or so.

Just one example as to who actually does the work on Fedora. A few days back I bumped into a strange memory leak in PostgreSQL libraries shipped with FC2. Posted a bug on Red Hat's Bugzilla for FC2. In the matter of hours, a Red Hat employee got on it and spent several hours attempting to duplicate the problem on his setup (admittedly with no success) and we left it off where I have to give it a try on an unspoilt FC2 machine (my suggestion) to verify that it isn't my own box that's causing the issue. So, this person at Red Hat actually provided several hours of unpaid support for me. I know that one cannot count on those things, but it nevertheless happens. It is Red Hat that make the majority of Fedora going, there is no doubt about that at all. Of course, other people help, but even if that help wasn't there, Fedora would still be released, IMHO. Possibly with more bugs, but released all the same.

The value of Fedora is in the delivery of the snapshots of a selection of state-of-the-art free software. Maybe not that great for production quality servers, but just fine for making sure you get a reasonably recent cut of binaries that will get timely updates when you need them. So far, Fedora has been relatively easy to keep upgraded, with minor issues related to software choice changes.

As for the stable kernel, I don't see how Red Hat's "stability" (or better, "backporting and enhancement") work cannot be incorporated into other distros. They only need to choose to do it, that's all. The problem is that Red Hat traditionally push it a little bit harder, mostly because they have clients that want to run Linux as a replacement for the big iron. Not to everyone's liking, but there is nothing stopping anyone from doing the same thing with Red Hat's code. There are absolutely no secrets there. In essence, this seems to be a political more than a technical issue.

I have to admit, I was sceptical about Fedora in the beginning, but it turned out better than expected, IMHO.

Posted Sep 9, 2004 12:50 UTC (Thu) by corbet (editor, #1) [Link]

"I am sort of concerned that now that Linus is not going to make stable kernels, the RH way of handling 2.4 might be the only way to do stable kernels. I would prefer to see more collective stability engineering take place."

Wow, Bruce, you understand things rather differently than I do. The whole point of the new kernel development model is to allow the distributors to stop backporting tons of stuff into older kernels, and to cut way back on the huge piles of distributor kernel patches in general. We are trying to get away from the "RH way".

Linus very much intends to release stable kernels; it's just that the way to get there has changed. The "don't allow substantive changes for a year" approach doesn't work very well. It's better to get the patches into the mainline and get them used. The result has been a lot more functionality in the hands of the users more quickly, and some quite stable kernels. 2.6.8.1 has not been the best example, but 2.6.7 was quite solid, and 2.6.9 should be as well.