Re: Threads docs updated in CVS

From: williamkempf_at_[hidden]
Date: 2001-08-12 08:05:29

--- In boost_at_y..., John Max Skaller <skaller_at_m...> wrote:
> Beman Dawes wrote:
> >
> >  >'xt is reached' ??
> 
> > Bill has already responded to this; it is on the issues list.   
Why it is
> > designed the way it is will be a FAQ, too, so we need to add 
something
> > there.
> 
> 	OK. My comment referred to the documentation, not the design.
> I assumed, naively, an elapsed time, but it is clear that a
> _particular_ time is the correct, now that it I am prompted to think
> about it: the reason is obvious--to schedule something 'precisely',
> you don't want to depend on when the call that makes the scheduling
> request is executed, you want to depend on an absolute time
> reference.  Question: what happens if the time has passed?

Actually, the compelling reason is that durations are near impossible 
to use within a loop that calls condition::wait().  You have to 
recalculate the duration on each loop, which will be expensive and 
not at all accurate.

> > Once we get a formal review (which I hope to post a schedule for 
tomorrow),
> > perhaps we can get some language experts from the Core Working 
Group to
> > help with the abstract machine.
> 
> Although even that is probably premature: we need users to try out
> the library to see how it stands up to a wide variety of stresses
> to gain confidence that the model serves well. Note that the
> next C++ Standard is still some way away, so we have time
> I think to do this right: it's probably the most critical change
> to C++ since Bjarne wrote the ARM.

Getting the wording "correct" as early as possible should be a goal, 
though.  We will have to correct things as we gain practical 
experience, but it will be easier to make such changes if the 
documentation is as precise as possible.
 
> Please note also that this abstract machine extension cannot be 
isolated
> to just C++. ISO C also requires threading support in the abstract 
> machine.

But they are going to be very different abstract machines.  We'll let 
the C folks deal with this if they have a mind to.

Bill Kempf