As the number of kernels threads increases (every thread seems to want one)
I wonder if we should switch to a different model.  Instead of having one
thread per "task", maybe we should adopt the common mechanism of having a
crew of worker threads ready to do work.

If higher level code want something to take place in a thread, it would
allocate a "work unit" and queue it to be executed.  (which would a wake
a waiting thread and let it dequeue the work unit and call the routine
indicated in).

-- 
Matt Thomas                     email: matt@3am-software.com
3am Software Foundry              www: http://3am-software.com/bio/matt/
Cupertino, CA              disclaimer: I avow all knowledge of this message.