global interpreter lock

[Bryan Olson]

Dennis Lee Bieber wrote:
 > 	Well, at that point, you could substitute "waiting on a queue" with
 > "waiting on a socket" and still have the same problem -- regardless of
 > the nature of the language/libraries for threading; it's a problem with
 > the design of the classes as applied to a threaded environment.

It's a problem for all levels; pretty much any of them can botch
it. Not only do we want to be able to wait on two queues or two
sockets, we might have good cause to wait on two queues *and*
two sockets. Win32 provides WaitMultipleObjects which lets the
programmer efficiently wait on many objects of various types at
one call; the Linux kernel Gurus are looking at supporting a
similar feature.

Library designers can botch the functionality by wrapping the
waitable object in their own classes with their own special wait
operation, and hiding the OS object.


-- 
--Bryan