[Ksummit-2009-discuss] Meeting userspace requirements

Ingo Molnar mingo at elte.hu
Sat Jul 11 12:34:55 PDT 2009


* James Bottomley <James.Bottomley at HansenPartnership.com> wrote:

> On Sat, 2009-07-11 at 19:32 +0100, Matthew Garrett wrote:

> > I think there's a perception that feature requests from 
> > userspace developers aren't treated with high priority - and 
> > "patches welcome" works less well in fields like the VFS where 
> > there's a lot of complexity. The O_REWRITE discussion seemed to 
> > end quite quickly even though everybody I spoke to thought it 
> > was worthwhile.
> 
> Actually part of the trouble with lkml is that the signal to noise 
> ratio is too low, plus there's a lot of, um, shall we say 
> arm-chair experts who like to opine on it, so people who might be 
> interested can just lose it in the noise.  O_REWRITE went through 
> so well mainly because it went to the fsdevel list not lkml.  In 
> general the subject specific lists tend to be far higher signal to 
> noise and far lower on the arm-chair opiner count.  We keep 
> discussing how to fix lkml to be higher signal to noise, but 
> no-one really has come up with a good solution.

There's a fair amount of noise on lkml indeed, but i have the 
opposite experience from you in the fields i'm interested in: 
specialistic lists often tend to produce annoyingly inbred, almost 
clan mentality driven code. Code from such lists often emerges in a 
too advanced state to really allow the resolution of design problems 
amicably once it hits a broader audience such as lkml.

lkml might be a busy and noisy place, but it is also refreshingly 
diverse. I find that the time i put into lkml is spent well, and the 
results (both for projects i input into lkml and projects i see on 
lkml) tend to be a lot broader and more reliable in the end than 
some of the stuff that comes from specialistic lists.

and the thing is, the most difficult technical problems in Linux 
tend to be cross-discipline - so them touching a generic list like 
lkml is an obvious bonus. Noise is a problem but can also be 
filtered out - on the other hand the lack of cohesion and the lack 
of a broader picture can cause irreversible loss.

	Ingo


More information about the Ksummit-2009-discuss mailing list