Re: [Nmh-workers] Locking (specifically, sequences)

[Ken Hornstein]

>i'm not asking for C11, though if we wanted MH to attract new users
>rather than just serving a declining/graying population, we would
>embrace the hell out of C11.

Really?  I kind of view those things are orthogonal.  I don't think
users really care too much about the language revision you're using;
they care about features provided.  Which admittedly we're not so great
on, but I think we're getting better.  We could switch nmh to F66 and
if we had improved MIME support I think everyone would be happy.  Well,
other than it might be hard to FIND a F66 compiler nowadays and we might
face a revolt just on principal from some of the greybeards here, but
you get the idea :-)  But if you think our reliance on C89 is holding
us back, please, elaborate!  I think we've pushed pretty hard on POSIX
as a minimum and I think that's been for the best for us.

I looked at the C11 Wikipedia page, and I do see a few things that might
help us (like improved Unicode support), but I don't think those are
showstoppers for any proposed work.

>with the quality and maturity of the MH core team, both in what they
>write and what they review, i am totally unconcerned about the usual
>problems that come with midlevel programmers encountering threads in
>their first large system. so, you're right in general, but who cares in
>this specific case?

Assuming I'm part of the MH core team (I'm not sure exactly WHO is on
it), thanks for the vote of confidence, Paul!  But I think on further
reflection I'll stick with the simple lock/sleep/retry cycle.  It just
seems like adding a dependency on threads just for this small bit of
code (which I suspect will rarely be invoked in the real world) just
seems like too much complexity for any possible efficiency gains.

--Ken