-> > Which brings me to my 2nd question: what sort of benchmarks would people
-> > be interested in seeing (if any)?  One of the main reasons to use AOLserver
-> > and PyWX is scalability; AOLserver is what AOL.com uses, so anecdotally
-> > it's pretty robust and scalable, but I don't know how to provide numbers
-> > demonstrating this.  "Hello, world" benchmarks are too artificial for
anyone
-> > to believe these days! Any suggestions?
->
-> I think "hello world" is still useful to get an idea of the application
-> server overhead.  I would be interested to see something that touched a
-> few megabytes of memory and used a second or two of CPU time.  It would
-> be interesting to see how this scales on an SMP machine.  Does PyWX use
-> OS level threads?

OK, I'll try it out.

AOLserver uses OS level threads, as does Python, and so PyWX uses them
as well; but you still suffer from the Python global interpreter lock,
so it will probably scale less well than Python embedded in Apache
(which has multiple processes and multiple Python instances, and so does
not suffer from the global interpreter lock).  Of course with Apache
you can't have a single database cache (or anything else that requires
centralized information) without jumping through hoops, for the same
reason.

Think of PyWX as a threaded FCGI system built into a single process ;).

--titus