So the use case is re-broadcasting messages from subscribers; it's kind of nice, and may simplify some patterns.
Gets my vote. :-) -Pieter On Mon, Dec 31, 2012 at 5:19 PM, John Muehlhausen <[email protected]> wrote: > To my knowledge, I have not used a sub with many pub connections, but I > assume that any subscription messages are copied to all of them. If that is > the case, then these upstream messages (that are not subscription related) > would seem to be copied to all as well, unless pains were taken to do > something different. > > As for a use-case, this one is contrived because I'm not sure that I am free > to talk about the real one. Imagine a chat system where every subscriber is > interested in a some or all of the messages, and where every subscriber can > generate messages. The new functionality could be used to get a new message > generated by a particular subscriber all the way back to the root publisher > (perhaps there are routing nodes) where it is then published to all who are > subscribing, perhaps with modifications (e.g., from a moderator). > > Without the modification there seem to be two approaches: > - use another socket, which doesn't seem ideal since we already have one > with upstream capability > - package the upstream "message" as a sub/unsub pair... I actually tried > this and the trie gets unwieldy above around 7000 bytes, particularly with > stack usage in the recursive trie routines > > -John > > On Mon, Dec 31, 2012 at 10:06 AM, Pieter Hintjens <[email protected]> wrote: >> >> On Sun, Dec 30, 2012 at 7:13 PM, John Muehlhausen <[email protected]> wrote: >> >> > Currently XSUB will not send messages unless they contain a subscribe or >> > unsubscribe prefix (see code fragment below). >> > >> > It would be useful to pass these "malformed" messages anyway, although >> > they >> > would not affect subscription status. >> >> It's an interesting proposal, but what specific use cases do you have? >> >> Would you round-robin if the sub has many pub connections, or would >> you copy to all connections? >> >> -Pieter >> >> _______________________________________________ >> zeromq-dev mailing list >> [email protected] >> http://lists.zeromq.org/mailman/listinfo/zeromq-dev > > > > _______________________________________________ > zeromq-dev mailing list > [email protected] > http://lists.zeromq.org/mailman/listinfo/zeromq-dev > _______________________________________________ zeromq-dev mailing list [email protected] http://lists.zeromq.org/mailman/listinfo/zeromq-dev
