From: Beman Dawes (beman_at_[hidden])
Date: 2000-06-01 09:16:24
At 12:45 PM 6/1/00 +0200, Branko wrote:
>Since I've been working on this for some time now, I decided to
>put it in the vault (folder synchronization). It's a first stab
>at defining an interface for basic synchronization primitives and
>synchronized operations. ....
Thanks for posting your code, but I think you may be missing the
point. Until we know what we are trying to achieve and what our
requirements are, there isn't any point in designing an interface.
The interface isn't the problem. The specifications are the problem.
The big picture that Dietmar and others have been discussing.
Think of it this way; people spun their wheels for years writing code
for containers. Then Stepanov came along with the
container-iterator-algorithm model coupled with the requirements
concept, and now the C++ community has a single, widely supported,
agreed upon way to do solve that problem. Good enough that others
abandoned their home grown containers.
What we (or at least I) am looking for is a multi-threading library
that is good enough to become a widely-used de facto standard. And
maybe even good enough to propose for the next version of C++
Standard Library. The only way that can happen is if the library is
based on a strong, theoretically correct, foundation. Let's have
that foundation before thinking about code.
It is possible that others such as POSIX or ACE or some authors have
already build a sufficiently strong foundation. If so, the then
point us to that foundation so we can decide for ourselves if that is
the direction we want to follow.
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk