Re: Review: Boost.Threads

From: williamkempf_at_[hidden]
Date: 2001-09-25 15:38:48

--- In boost_at_y..., "David Abrahams" <david.abrahams_at_r...> wrote:
> I hate to just stand on the sidelines and declare "hear, hear!", 
but,
> well...
> 
>    "hear hear!"
> 
> I also found Peter's arguments quite compelling. This speaks well 
of the
> level of discourse we have here.
> 
> I realize that the problems Bill K. has had to face in designing 
this
> library were/are substantial, and there is definitely some value in 
limiting
> the scope of any problem we try to solve at one time. Maybe we just 
have to
> say, "as a first step, we can't expect to come up with ``good'' 
names and
> easy semantics right away".
> 
> All the same, it is never satisfying to have to lower one's 
standards, and
> we know we can do better in the long run. At boost, however, we are 
all
> volunteers. It behooves us to acknowledge the impressive job Bill 
has done,
> much of which was the research that allows him to engage with 
authority in
> debates over the library's design. Part of that acknowledgement, I 
think,
> has to be the willingness to pick up the ball at this point. If we 
are going
> to server community (2) below, some of us may need to begin the 
design of an
> appropriate interface layer.

A big word of warning... such a layer should not be ANYTHING at all 
like these primitives.  There are just too many pitfalls for new 
users to base an interface off of them.  This was (and is) the goal 
of Boost.Threads.  The current library provides us the basis for 
development of higher order concepts such as CSP concepts.  It 
provides the portable layer needed for such higher order concepts, 
and has the added benefit of being usable by "experts" (or those 
willing to take on the very steep learning curve) for MT programming 
with out the higher order concepts.  The target audience of this 
initial version of Boost.Threads is not for newbies.

Bill Kempf