This will require establishing vector clocks between your nodes and still requires a conflict resolution protocol of some sort to decide what to do when updates with the same timestamp conflict, etc. If it were as simple as adding a timestamp then distributed database transactions would not be considered a "difficult" task. Again, if ordering does not matter then this problem is pretty simple (and timestamps may be a useful addition to the data payload) and if ordering does matter then this problem is very difficult and some parts of it might be impossible to solve at the DHT level.
jim On Mar 6, 2008, at 10:45 AM, Marco Milanesio wrote: > Hi, > You can try to put some other information on the data load, i.e., a > timestamp for versioning the > elements of your list, and of course, modify the primitives of the > DHT. > When a newer content is inserted, at the application level its > reference > is inserted into a list, which is > the *value* of the key, and then the whole object is republished. > > Marco _______________________________________________ p2p-hackers mailing list [email protected] http://lists.zooko.com/mailman/listinfo/p2p-hackers
