Dave Cridland typeth: | more practical. Indeed, having chains of sub-MUCs becomes practical, | since either the MUC is public, or potentially a chain can be set up | with bilateral agreements, which is effectively how IRC and PSYC | operate.
Actually no, there is quite a difference in how conference control is designed in the two technologies. With IRC each server has full conference control over a channel on the network - each one of them can say XY is an operator and YZ is banned. In the case of netsplits there are various logics to figure out which side is authoritative, but if the server is evil, it can always say its users started that channel somewhere around 1970 and thus have the authority over the channel. PSYC has a context master (theoretically more than one, but we didn't implement any of that). One server hosting the context program which defines the conference control for the chatroom. All nodes that are involved in distributing the context to its members may be entitled to perform certain jobs on his behalf, but the protocol always makes it clear who has the authority. That is quite a dramatic difference. By having full MUC implementations interconnect it sounds a bit like there is a chance you'll run into the authority problems IRC has with its distributed conference control. By taking the distribution efficiency problem deep into details of MUC and pubsub you are looking into opening large cans of worms instead of keeping things simple and providing for a fundamental one-to-many routing plan first, methinks.
