KS2007: The distributor panel

KS2007: The distributor panel

This is precisely correct. This is the real problem. The Linux devs no longer have reliability as the central, core goal. It's A goal, but third- or fourth-tier... the primary goal is writing new code, making it a fun thing for developers to play with. When it went to the fast-release cycle, they declared their permanent derision for end users, the people who have to actually use this stuff in real life. In real life, stuff has to work, and the only way to make sure it does is with testing and slow development. A two- or three-month release cycle is a gigantic middle finger to the people that need to trust the code.

When the 2.6 kernels started to fall to shit, a lot of us complained. We were told that we weren't supposed to run kernel.org kernels, and and that the distributions would take care of testing and bugfixing. Instead of actually DEALING with the problem and shipping reliable code, they waved their hands in the air and expected others to handle it for them.

Well, guess what? Others are indeed handling it, but not how the poor kernel devs want. Awwww. The way they've found to deal with it is by not running kernel.org code; it's by using older releases and very, very carefully thinking about what to backport. The kernel people are mad because they got exactly what they asked for. The distributions are bugfixing and providing their customers a product that can actually be used.

Nobody in their RIGHT MINDS runs code straight from Linus anymore, because it's a seething, bug-infested mass of crap. The release cycle is ridiculously fast, the code quality has fallen to shit, and nobody can even be bothered to fix bugs.

And they DARE to complain that people aren't using their most recent product? This is insane. Ship code that WORKS and people will use it. Stuffing crap in a crate and expecting it to turn into gold by lots of user testing just infuriates your users, whose jobs depend on this garbage working. Every time.

Reliability is what made Linux successful in the first place; you could trust the 2.2 kernels with your life. Early 2.4 was terrible, and never got as good as 2.2, but by late in that cycle it was quite robust. 2.6 has been a steaming bag of crap.

Speed of change is not a useful metric of progress. Speed of RELIABLE change might be, but that's going to take a total shakeup of the broken dev process in place now.