[OpenAFS] OpenAFS on Linux 2.5.x

[Russ Allbery]

Steven Jenkins <steven.jenkins@ieee.org> writes:

> If it's really true that linux-kernel is *hostile* to the real-world
> context---how their code is used, and by whom, and for what
> purpose--then I think they will eventually make themselves irrelvant.

It's Linus's belief that anyone with a need that they want Linux to meet
will care enough about it to write the code well enough that it will pass
technical muster, and if people can't do that, he doesn't worry about
supporting them.  This development strategy does work surprisingly well,
although the determination of "technical muster" invariably involves some
subjectivity and there are places where I'd flatly disagree with his
determinations.  (For example, the crippled AFS client he accepted into
the kernel does not even remotely pass technical muster IMO.)  But he's
Linus and I'm not, and it's hard to argue with success.

> But I really don't want to carry on too much about this one issue. The
> real question is: do organizations with a major stake in the future of
> AFS feel any need to act in concert to advocate for AFS, or are you
> comfortable with what we have now? If we need to advocate, what should
> we advocate, and to whom?

Well, one place to start would be to advocate amongst each other to ensure
that people are aware of the issues.  For example, I saw an extended
dust-up about PAGs in the kernel and whatnot, but I didn't glean from that
a lot of detail about what the actual problems are or why it doesn't look
like Linux 2.6 will work with AFS without patches.  So a good place to
start would be some sort of summary of exactly what the problem is and why
it's not yet been solved.

If I had a summary, I could take that to my management and we could bring
it up in discussions with Red Hat and make it clear to Red Hat that this
is a very important feature for us (if that will do any good, which is
unclear, but it can't hurt).

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>