Re: [boost] RFC: Multithreading design constraints

From: Beman Dawes (beman_at_[hidden])
Date: 2000-06-05 10:48:59

At 07:31 AM 6/5/00 -0400, John Maddock wrote:

>I've been following the comments on multi-threading recently with
interest,
>as it's something that's important to me, I've put together a set of
design
>principles (below) that could serve as a "straw man list" starting
point

Interesting reading, to say the least.  But there are meta-questions
we need to address before charging off and developing our own set of
definitions.

Do we want something that is suitable for ISO standardization?

Do we want to start from scratch or do we want to build on work
already done?

There is already an ISO standard that covers multithreading (ISO/IEC
9945:1-1996 POSIX).  We ought to explore the possibility of basing
our work at least partially on that standard.  (Background: ISO is
unlikely to approve any new C++ standard for multithreading that
totally ignores another ISO standard for multithreading, particularly
one that includes a C language API.) 

I'm not suggesting we include the POSIX pthreads C API, but rather
use their definitions where applicable.

It would also save a tremendous amount of work to specify any boost
functionality as if implemented via an ISO/pthreads subset.  One
possible subset would be the GNU pthreads/Win32 project; that would
tend to ensure portability across at least UNIX and Win32
implementations.

It is fun to go off and do our own thing, but we also ought to
leverage the many years of work other groups have put into
multithreading.

--Beman

PS: ISO/IEC 9945:1-1996 POSIX costs $ 245 US, and is not available
publicly in electronic form.  Thus we would have to extract
definitions, and would probably use something like David Butenhof's
"Programming with POSIX Threads" as a working reference.