> On Dec 25, 2006, at 10:10 PM, John Cowan wrote: > > > Shawn Rutledge scripsit: > > > >> What other kinds of constructs can you imagine in Scheme for > >> parallel > >> execution, which would work better than threads? > > > > Communicating sequential processes (CSP) would be the obvious > > candidate. > > All the pieces for a Chicken Termite are in place: > > match-action egg > mailbox egg (w/ timeout & cursor) > Thomas Chust's rpc egg > logging egg > remote-launch egg
Even more: www.askemos.org has a slightly out of date chicken tarball. That one already served non-mutating requests before. Background: - we plan to go multi plattform anyway, but it takes time - the chicken based dev version is stalled for performance reasons (at chicken version 0.something) - Askemos/BALL is slightly more advanced at the application level (e.g. we don't ask in the manual to stay away from side effects, we avoid them in the interpreter); otherwise it's highly simillar - it doesn't have message matching (at application level) since it's are completely unclear how this could interfer with the synchronisation protocol - it has a byzantine sync protocol, there is global shared state (sort of "software transactional memory") over a quorum of hosts (the quorum is a per object property) - the shared state is beyond the exclusive control of individual machines (required for tamper proof processes like electronic contracts, where Askemos is currently the only know option which actually works) - porting requires mostly pulling eggs (is there something like libmagic and iconv?), convert FFI and some C in parsing and serialising - we would have done it long ago, we just don't have the time. Money will probably bring us even a Java version before the chicken port even though the Java port will be much more than a few weeks. _______________________________________________ Chicken-users mailing list [email protected] http://lists.nongnu.org/mailman/listinfo/chicken-users
