I agree that this can be a problem when teams perform equally well. But it
does not seem easy to change the RCRSS in such a way that messages are
always guaranteed to arrive in the same order. This seems to be the price
of running agents & simulators indepently of the kernel. Although I think
this behaviour can be `fixed' this will probably involve major changes.
It is very likely that this problem really exists since the simulators can
only send SK_UPDATE messages if they have finished their internal
execution for a particular simulation cycle. Sometimes simulator B may
finish execution before simulator A, resulting in a different ordering of
SK_UPDATES to the kernel. This is how I see things, but I'm not entirely
sure if this is all strictly true.
This archive was generated by hypermail 2.1.3 : Tue 16 Dec 2003 - 16:57:46 GMT