Blaine Cook wrote: > On Sat, Apr 12, 2008 at 12:23 PM, Nick Vidal <[EMAIL PROTECTED]> wrote: >>> 1. What do web APIs look like when translated to XMPP, and what new >>> interaction models does that enable? >> We could very much relate AtomPub with PubSub. These two standards enable >> users and agents to discover, subscribe, view and syndicate information in >> an intuitive and non-obtrusive way. > > The twitter api actually does this, albeit unofficially and with some > warts. Alex has begun documenting the public timeline pubsub feed > here: > > http://groups.google.com/group/twitter-development-talk/web/jabber-pubsub > > but so far unreleased (but complete) is the actual PubSub (XEP-0060) > API for subscribing to feeds. Authentication and interface design are > the main blocking issues preventing Twitter from having this in the > public. > > FWIW, I don't think *publish* over XMPP makes sense in the context of > Twitter until federation becomes a reality, since the design of > clients is simplified by just using HTTP.
Yes I've thought about renaming the Atom-Over-XMPP spec "AtomSub". You can publish however you want (AtomPub or some other technology), but the notifications are best pushed out via XMPP (IMHO). >>>> 2. Federated social networks over XMPP >> I think ISS (Instant Syndicating Standards) provides a good framework >> towards federated social networks over XMPP and HTTP. It's worth taking a >> look. > > Agreed - I must confess I'm not excited by the tag clouds / discovery, > since I'm a big believer in simplest-possible-thing. HTTP has managed > to get by with 3rd party discovery mechanisms that are built on top of > the protocol layer (google, yahoo, etc) rather than inside the > protocol layer. But in general, I think the patterns are sound and the > key is the addressability, which XMPP provides where HTTP doesn't (in > a guaranteed consistent way). Speaking of which we need to standardize a <link rel=''/> for XMPP nodes. Ralph, what were the results of your conversation with Dan Brickley about the 'type' attribute? Shall we just leave that off? Peter -- Peter Saint-Andre https://stpeter.im/
smime.p7s
Description: S/MIME Cryptographic Signature
