The final session of the 2003 Kernel Developers' Summit consisted of a look
back at the 2.5 development series and a look forward to how 2.6 and 2.7
would work. Ted Ts'o started the session off by noting a rare and pleasant
event: Linus turned down a patch - even though he liked it - because the
focus is currently on stabilization. Ted noted that 2.3 did not go so
well; Linus is getting better at saying "no."
Linus talked about one thing that worries him: the linux-kernel list. It
works well from a technical standpoint, but it has become, at times, so
hostile that people are afraid to post there. Linus says he used to be
amused by linux-kernel flaming ("Al Viro is a work of art"), but that it is
no longer funny. He asked the kernel developers to be aware of how they
treat people on linux-kernel and try not to push people away.
It was asked whether linux-kernel should be split into separate lists for
the stable and development kernels. It was noted, though, that it is
already hard enough to keep patches for the two branches in sync;
separating the discussions would make things worse.
Daniel Phillips noted one thing that went very right with 2.5: Andrew
Morton. The comment was accompanied by general applause.
One thing that is not right is that many drivers - especially out-of-tree
vendor drivers - have not been ported to 2.6. The suggestion was made that
2.6.0 should be released soon - even if it's not entirely stable - so that
the vendors would pay attention to it. The sense seemed to be, however,
that an unstable 2.6.0 would not help the situation at all.
A good thing in 2.5 was the adoption of BitKeeper, which made things better
even for the developers who do not use it.
A not-so-good thing is the large volume of 2.4 patches which still have not
made it into 2.6. Linus asked that important patches - security patches in
particular - be marked so that they don't get dropped. Dave Jones, who
worked on bringing 2.4 patches forward for a while, noted that the job gets
overwhelming after a while and really requires more than one person to do
One more good thing was the "what to expect" document maintained by Dave
For 2.7, Linus noted that there really are not a whole lot of pressing
issues at this time. "It's kind of scary, but it's good." Most of the
developers are not too worried about finding things to do.
Ted Ts'o noted the long release cycles in 2.4, and asked whether things
couldn't be tightened up, complete with "feature freezes," for 2.6. Andrew
thought that could probably be done. Andrew also stated that he wanted to
be able to get more ambitious changes into 2.6 than have made it into 2.4
so far. In response to a question from Linus, though, he noted that
replacing the entire VM subsystem in 2.6.9 is not in his plans.
to post comments)