KevWould people think it unexpected if they did a disco#items on pubsub.server.tld, and got back results for email@example.com, firstname.lastname@example.org, email@example.com... each with a single node per JID?
Kev(Note use of 'unexpected' rather than 'illegal')
ralphmKev: I think so, yet.
ralphmIt doesn't seem illegal, though.
ralphmReading the prose in XEP-0030, it does seem to favor addressable entities (i.e. things with a JID) over using nodes.
ralphmHaving an empty node identifier is ill-defined in XEP-0060. Of course there's a reference to a section on collections, but that section has moved to it's own XEP.
ralphm(and I am still of the opinion that collections are not a thing to want, even though it might appear that way at first)
dwdralphm, I like collections.
ralphmdwd: no you don't, trust me.
dwdKev, Unexpected, yes.
dwdralphm, I implemented "full" collections. I really quite like the notion.
ralphmI haven't found any compelling use case where you want to deal with DAGs and all the explicitness of setting them up. I did think I wanted to use them on multiple occasions, and then always did a second implementation with implicit nodes as an interface to an existing (or new even) backend's business logic.
ralphmSome of those nodes would kinda feel like collections, but without all the hassle and fixed semantics.
dwdYes, I think if you're implementing a customer pubsub layer over existing logic, you don't need them. But I'd prefer a way to avoid that being the only way to deploy pubsub.
ralphmdwd: I'm sympathetic to that idea, but it always seems to result in distributed knowledge about 'resources', with all the synchronisation issues that come with that.
ralphmI'd like to compare this with the way the web currently works, especially after CGI was 'invented'. Most of the time, you don't explicitly create resources there, either.
Steffen Larsenhas left
dwdDon't *say* that. The REST people will crawl out of the woodwork.
dwdOnce they start, they never shut up.
ralphmFortunately, I'm room admin.
ralphmAnd I'm also ok with creating leaf nodes explicitly. But collections are arguably designed to mimick some existing backend topology, and I think it's better to do that by making the pubsub protocol part of that backend, instead of treating it as an external, second-class thing.
ralphmTaking that to its conclusion, you generally don't need all the explicitness of collection node configuration.
ralphmAnd probably also not exposing the internal structure of where an event came from.
Kevralphm / dwd: Thanks.
ralphmKev: I'd love to hear more about the actual use case.
KevI think it's just a wart of someone not knowing what the usual way of doing things was.
KevI saw this happening, there's no reason for it to need to be this way. I just wanted to check I wasn't just being Wrong before I suggested it was unusual for them to do this.
dwdAs I recall, at one point, that was considered the right way.
ralphmdwd: wait what?
dwdralphm, Pubsub nodes having a local-part to their jid.
KevI don't remember this.
ralphmdwd: I think you are confused with http://xmpp.org/extensions/xep-0024.html
dwdOh, no, I'm muddled. It was resource-parts that were added.
dwd'60§4.6.1 I'm thinking of.
dwdThe Node identifier MUST NOT be specified by the node identifier, in fact.
dwdYay for overloaded terms!
KevAmbiguity is underrated.
KevAnd, to be clear, when I say 'ambiguity', I'm using it in the sense of 'being clear'.
ralphmKev: if you want to argue against 'wrong', have a look here: http://howtosdeit.blogspot.nl/2014/02/comunicate-rabbitmq-with-publish.html?spref=tw
KevAs an aside, I've decided not to put up Swift ideas for GSoC this year. I don't think I've got time to mentor properly.
ralphmKev: rather that then finding out while students are at it.
ralphmKev: good call
KevIt's a shame. I like GSoC, but work has become...busy.
ralphm"They say Google is dropping #XMPP on May 15th. So it seems we are dropping Hangouts in the day of our first anniversary. F*ck them!" — LoquiIM (https://twitter.com/LoquiIM/statuses/432867870016434176)
Tobiasare they referring to some magic 8-ball?
ralphmDon't know. But if that turns out to be true, I can see how they will support unauthenticated encryption on all s2s traffic by May 19.
SimonAnyone know any of the Loqui guys?
ralphmI don't think I do.
SimonI've asked Sonny Piers since he works in Spain and in XMPP and I don't think there are many XMPP startups in Spain.
ralphmAs a counter argument, I know virtually no XMPP developers in The Netherlands, where I'm sure there are some.
dwdI mean for the May 15th thing.
dwdAh, Google Voice is dropping XMPP support on May 15th.
ZashGoogle Voice has XMPP support?
dwdhttp://www.androidpolice.com/2013/11/04/merging-of-google-voice-and-hangouts-will-result-shutting-down-all-3rd-party-voice-apps-in-may-2014/ and so on. Announced on 2013/11/01 I think.
TobiasZash, come on, admit it, you'll gonna miss it
ZashYes, me, of all people, is gonig to miss it.
Link Mauvehas joined
Link Mauvehas joined
ralphmdwd: yeah, that's the only stuff I could find, too. I'd be very amazed if all XMPP support, i.e. the old Google Talk network, would be shutdown at that date. Of course I've been surprised before.