On Sun, Jul 25, 2010 at 10:21 AM, Oliver Smith <[email protected]> wrote:
> Before a given client can start applying changes, though, it needs the > most recent snapshot of the overall picture. > > My first-hash implementation adds a thread listening on a second socket > REQ/REP socket, which builds an image of the current snapshot and then > itself joins the PUB/SUB socket. You may still get race conditions unless you subscribe at the same time as you request the snapshot, queue updates, and then apply them to the snapshot. Presumably the snapshot is very large compared to the updates? What is your strategy for recovering from loss of some updates? -Pieter _______________________________________________ zeromq-dev mailing list [email protected] http://lists.zeromq.org/mailman/listinfo/zeromq-dev
