[arch-dev-public] Kernel - vanilla vs patched?

[Paul Mattal]

Thomas Bächler wrote:
> Tobias Powalowski schrieb:
>> it's not about vanilla and or not vanilla it *HAS TO WORK* and must be 
>> maintainable in a reasonable amount of time, thats the  real thing of a 
>> kernel.
> 
> If you would have read my mail, you knew that this is exactly what I
> have doubts about: How do you know that a certain bugfix works and
> doesn't add even more bugs?

You can't be sure. All you can do is make small modifications that 
you believe you have your head around, and try to do no harm. And 
fix/revert it if it breaks something.

But I don't think we should look at something obviously wrong and 
say "no, we're going to leave that broken". It's a nice ideal, but 
it's just not practical, like Tpowa says.

It's very easy to not apply bugfixes, until the day when Linus's 
kernel break your network driver. Then it's just a hassle for 1000 
people to have to recompile the kernel package because we won't 
include the fix.

We make things work, and we accommodate the needs of our community, 
as cleanly as possible. I think we should continue to do that.

Per Jason's fine example on the CVS stuff, I'm going to try to pick 
something relevant to the situation and actually do it to improve 
things. I'm adding to my list to try to get unionfs to work outside 
the kernel. It may take a while, possibly even weeks, but I'm 
choosing that one and I'm going to work on it. In the end, it will 
probably need some in-kernel patches to expose functionality, but I 
will try to reduce them to the minimum.

I will report back here on my progress.

- P