jarl van katwijk wrote: > > Ok, we wont go for the all-on-one-place design, that's for sure. The central > service I have in mind is just for locating other piper systems. Nothing else, > just a list of IP numbers. All other stuff is P2P. I thought about this for a while: How will a user know who they can connect to (who is out there) when there will be no central directory service? And I think the solution is simple. We can do what XChat does: include a list of addresses with the software. Perhaps we can have a modifiable list for each user, let's say an "address book". We'll simply put a few addresses in there to start them off. Actually, if we have a few "big nodes" (or instances of Piper) running (at Bioinformatics.org for one), all that the users will have to do is connect to them, and, because Piper is a P2P system, users will have access to more than what is local to the big node. Users will have access to all instances of Piper connected (all that are public anyway). The effect would be nearly the same as having a central directory service. > There're two thing here: transport and language. I wasn't even considering the > transport until now ;) But after the bad news about the dialects of KQML I found > this: http://www.fipa.org/specs/pesspecs.tar.gz. I'll try to isolate relevant > documents\pages and make them subject to discussion. If we're thinking of Piper-to-Piper communication as something like agent-to-agent communication, perhaps there is something we can share with Narval? But, I don't know, maybe we don't want to get into SOAP when we are already using CORBA. Cheers. Jeff -- J.W. Bizzaro jeff at bioinformatics.org Director, Bioinformatics.org: The Open Lab http://bioinformatics.org/~jeff "All those scientists--they're all alike! They say they're working for us, but what they really want is to rule the world!" -- Angry Villager, Young Frankenstein --