new beta: 1.6b5

Hi. I managed to get a bit of coding done finally, so here is 1.6b5.
This version has several experimental features.

The buffer size is dynamically adjusted to be a multiple of the TCP 
segment size. This should increase performance with small buffer sizes. 
(Well, hopefully.)

Another enhancement is the addition of the NumProcesses keyword. If the 
value is greater than one, mathopd will split into several parallel 
processes. The idea behind this is to take advantage of multi-CPU systems. 
Since I don't have a MP machine, I would appreciate it if anyone can see 
if this actually has any benefit in real life.

Cheers
Michiel