Standards Update, USENIX Standards Watchdog Committee

[Peter da Silva]

From: peter at ficc.uu.net (Peter da Silva)

I've finally got a copy of P1003.4, and I find it to be quite nice. The
lack of threads is no big deal... threads should certainly be standardised,
but any threads design that can't be implemented on top of P1003.4 is
probably going to cause big problems for existing systems anyway.

One thing to consider is that threads and real-time are not equivalent
concepts. Threads are a nice technique for implementing real-time systems,
and most real-time systems make an implementation of threads pretty easy,
but there are non-real-time systems that implement lightweight processes for
reasons of improving throughput rather than reducing response time.

Keeping P1003.4 from prohibiting certain threaded implementations is one
thing, but it shouldn't require threads in any real-time system. And it
shouldn't require that you have to go to a real-time system to conform
to the threads standard.

Threads probably deserves a P1003 number of its own.

As for Berkeley's sore feelings because P1003.4 doesn't look like BSD, that's
just silly. It'd be like USG being upset because P1003.4 doesn't implement
the System-V IPC kludges. P1003.4 looks quite familiar to me, from working
with other real-time systems... including real-time-like UNIX. And it should
be implementable (as far as the functionality you need for real-time can be)
on top of sockets, without penalising real real time folks by sticking them
with a socket interface.
-- 
 _--_|\  `-_-' Peter da Silva. +1 713 274 5180.      <peter at ficc.uu.net>
/      \  'U`  Have you hugged your wolf today?  <peter at sugar.hackercorp.com>
\_.--._/       Disclaimer: commercial solicitation by email to this address
      v                    is acceptable.


Volume-Number: Volume 19, Number 99