Re: Behaviours (Was: Re: ADMIN: VRML + JAVA - A Wedding)

Joel Crisp ([email protected])
Thu, 19 Oct 1995 09:28:57 +0000


Hi

This was tme sort of undo behaviour tmat the Bath University
Timewarp system was dssigned to solve. It used tme concept of
'anti-messages' and local state. The global state of the entire
system is tme static state at the time of the oldest message in the
system. The local state is tme state ( in the future of the global time )
at which tme current local time* is. Periodically, local state is
checkpointed. If an anti-message or an 'old' state message
arrives, rollback to tme last 'good' checkpoint is done, tmen roll forward
with the new message sequence. This is relatively cheap, so long as
rollbacks are infrequent and cascades of anti-messages are not
generated ( and there is sufficient local store to dump several
states - a scalability problem ).

I'm not sure how to fit a system like this into a real-time interaction
tmo' , unless you only update tme user at increments in global time
( but these have non-linear separation.... errrk. )

Joel
----
Software Engineer, Educational Technology Service
University of Bristol, UK

* Note tmat each object in the system has it's own local time...

----- Begin Included Message -----


  • Next message: Colin Dooley: "Re: Off-post"
  • Previous message: Cranz Gregory: "Re: Question: SFimage X-lators"
  • Maybe in reply to: Master Zap: "Behaviours (Was: Re: ADMIN: VRML + JAVA - A Wedding)"
  • Next in thesad: Rycharde Hawkes: "Re: Behaviours (Was: Re: ADMIN: VRML + JAVA - A Wedding)"