Kevin Smith wrote: > On 27 Aug 2007, at 17:28, Peter Saint-Andre wrote: >> We had a long long discussion thread about this a few months ago, as a >> result of which we modified rfc3920bis to recommend the use of random >> resource identifiers that are generated by the server, not the client. > > FWIW, I don't agree with the notion that these random resources are a > good thing
Yes, you have previously expressed that opinion. :) > (I'd rather go back to the 'dark days' where we could infer > meaning if we wanted to (albeit incorrectly, by rfc)). Well it's another SHOULD-path vs. MAY-path discussion, isn't it? We recommend that you ask your server to generate the resource identifier but say that optionally you can generate one client-side if you want. > I think the solution is simple though; if the server isn't routing your > presence to someone, it should reply to iqs on your behalf saying you're > not there. This is consistent with the route people have been suggesting > recently (and I think I agree with) of 'if you want to start an X > session with someone not on your roster, send directed presence first'. I'm not yet sure that's the right approach. I mean, it seems quite reasonable to me (which is why I keep mentioning it), but Ian Paterson has objected that you might want to engage in a stanza session with someone but not share presence (e.g., a Jingle call). Against that I mention the traditionally quid-pro-quo nature of IM systems (give out your presence, engage in real-time communication). As I say, presence sharing (even temporarily for a transient chat) seems like a reasonable expectation to me, but it might not seem that way to the participants in certain cases. Peter -- Peter Saint-Andre https://stpeter.im/
smime.p7s
Description: S/MIME Cryptographic Signature
