[Ksummit-2009-discuss] Meeting userspace requirements

Matthew Wilcox matthew at wil.cx
Sat Jul 11 11:53:48 PDT 2009


On Sat, Jul 11, 2009 at 08:02:48PM +0200, Johannes Berg wrote:
> That would be my first guess too -- no offence meant to the regular lkml
> folks (some people really do irk even active contributors), but it seems
> to me that new people who have actual _requirements_ to offer will just
> be shot down 99% of the time because you need to be able to read between
> the lines of their statements, not take their statement word for word,
> and often they'll need coaching on how to achieve a userspace solution
> instead of just getting "this need not be in the kernel period".

Possibly we need some kind of document on interacting with the kernel
community (which actually isn't that different from interacting with any
other technical community).  To wit:

 - Bring problems, not necessarily solutions
 - Be willing to investigate, try, and quantify different solutions
 - Understand that we aren't willing to completely change the kernel to
   suit their application, but we are willing to make changes that we deem
   "architecturally clean".
 - What do you hope to gain from kernel assistance?
 - Why are existing kernel facilities inadequate?


I think we have some interesting case studies, one of which Jon has
expertly analysed: http://lwn.net/Articles/339316/

Other examples might include AIO, performance monitoring, NPTL vs NGTL,
futexes, inotify/dnotify/fanotify, ...

Lots to think about, and I bet we could easily chew up an entire session
at KS talking about our successes and failures at implementing kernel
features which required userspace to take advantage of them.

-- 
Matthew Wilcox				Intel Open Source Technology Centre
"Bill, look, we understand that you're interested in selling us this
operating system, but compare it to ours.  We can't possibly take such
a retrograde step."


More information about the Ksummit-2009-discuss mailing list