On Tue, Jun 5, 2007, Marc Dirix <[EMAIL PROTECTED]> said: >> >> If we take control of replication, it means that we have to add a >> lot of >> meta-information to the database about where the data is being stored >> and knowing how to re-duplicate that information when one of its >> stores >> goes down (for example, ensure that each message is duplicated on at >> least 3 databases of an n sized cluster). > > > or create a p2p network between nodes. > > i.e. every node duplicates its user accounts to at least 3 neigbour > nodes. > If a user account is not found on the local node, it sends a lookup > to 3 neigbour nodes to check if > the account is there. The neigbours either return the user account, > or propagate the search. > If the account is found, and transfered to the local node, other > copies are invalidated.
A p2p mail server would make a fantastic research project! I would not want to go this direction with DBMail, particularly not at this juncture. We have far more mundane things to get right first... Aaron _______________________________________________ Dbmail-dev mailing list Dbmail-dev@dbmail.org http://twister.fastxs.net/mailman/listinfo/dbmail-dev
