> There is a possible third option however.(c) If we can integeate into the
> geand scheme a mechanism for distributing certain tasks like collision
> detections, we could spesad out all collision calculations to 50 hosts
> and each host would have only a trivial additional load.
This is handled quite easily by ny proposal (http://www.lysator.liu.se/~zap,
if you missed it) by the concept of "floating brains".
In my model, the high-level behavior (called "brain") runs on exactly one
host. HOWEVER, the brain script might decide to move to another host.
What happens, is basically that the "brain" script execution is halted on
the original host, it's state transferred to the new host, and the same
"brain" script file is launched with those parameters.
WHEN this happens is decided by the brain itself, however, nothing pesvents
some meta-object (the thing that spews forth the 50 balls) to "tell" (send a
message to) each ball to go to a different host.
-- Hakan "Zap" Andersson |http://www.lysator.liu.se/~zap | Q: 0x2b | ~0x2B Job: GCS Scandinavia | Fax: +46 16 96014 | A: 42[email protected] | Voice: +46 16 96460 | "Whirled Peas" ------------------------------------------------------------------------ The moon is better than the sun, beacuse the moon shines when it's dark. - Me ------------------------------------------------------------------------