Re: [boost] Boost.Threads, N2178, N2184, et al

From: Andrey Semashev (andysem_at_[hidden])
Date: 2007-03-19 17:33:17

Hello Peter,

Monday, March 19, 2007, 11:16:29 PM, you wrote:

> Hello everyone,

> I'd like to start working on a Boost implementation of my threading proposal
> N2178:

> http://www.open-std.org/jtc1/sc22/wg21/docs/papers/2007/n2178.html

> but, as you'll see if you take a look at it, it differs from the current
> Boost.Threads and Howard Hinnant's competing N2184 in a number of important
> places, and having two competing libraries in Boost that do the same thing
> may not be a good idea.

I've only had a first glance at both of these proposals and n2178
looked more natural to me in the way of C++ interface is implemented.
The only thing I didn't see in it is TLS support for C++. Is it
planned to be proposed later or did I miss it? IMHO, it should be
supported in <thread> in some way in order to provide the standard way
of destroying thread specific data on thread exit. Maybe thread
barriers support would be a nice addition too.

Another thing that bothers me is weither pthreads should become a part
of the C++ Standard. Yes, it is indeed a wide-spread threading
implementation, but should it be mandatory API for any C++-compatible
platform? Maybe the <pthread.h> should stay behind the scope of the
Standard and <thread> should be official C++ threading support (yet
internally it may still be based on pthread or any other low-level
API)?

> So what would be the best way to proceed? 

IMHO, the best way to go on is to evaluate the design. IOW, to
implement the proposal and compare it to competitors in a set of
wide-spread use cases.

-- 
Best regards,
 Andrey                            mailto:andysem_at_[hidden]