Hi Trejkaz, On Sat, Aug 27, 2005 at 08:56:15PM +1000, Trejkaz wrote: > On Sat, 27 Aug 2005 20:51, Rory wrote: > > The current situation - > > 'gmail.com' domain hosted on 'talk.google.com' without > > the DNS SRV record - requires unnecessary non-standard > > (as per RFC3920) additions to an XMPP client/library. > > Something like connecting to a different address isn't exactly rocket > science, > though. Naturally it would be good for them to have these records to make > users' life easier, but having to connect to a different IP address is > something that most libraries have to implement anyway, and that goes for any > protocol. :-) >
I understand that it is easy, I've already made the modifications to my library to accomodate it, I just don't want to make those modifications permanent because they are unnecessary cruft. If the spec is followed then there is only a need to accomodate the following information from the user/automated client: (i) their JID; (ii) their password or other authentication token. As far as I'm concerned this is nice, clean and the way things should be. If you permit the user to provide you with a server name and port number then things become more complicated than they need to be. For instance, does the user-supplied server name qualify as a valid identity for the purpose of validating the server's X590 certificate. Or, if there this is an automated client, should we fall back to checking for DNS SRV records if there is no server listening at the specified server address? Do we offer the user configuration options to answer these questions and the others that arise? ... I guess I find that the RFC3920 is flexible and complex enough that I don't wish to see additional de-facto standards materialise that need to be supported. Rory _______________________________________________ jdev mailing list [email protected] http://mail.jabber.org/mailman/listinfo/jdev
