2009/10/3 Peter Saint-Andre <[email protected]>: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > On 10/2/09 6:26 AM, Dave Cridland wrote: >> On Fri Oct 2 12:44:21 2009, Matthew Wild wrote: >>> Definitely. I'd even go so far as *recommending* a default path, or at >>> least vendors coming to some informal agreement. Currently we have >>> /http-bind, /bosh and /xmpp-httpbind and possibly others I've missed - >>> I'd like to settle on /bosh. Perhaps I'm just being too optimistic :) >> >> I think you may be - I vaguely recall an endless discussion about >> well-known HTTP URI paths within the IETF, and it wasn't pretty. > > Oh yes, there was a big discussion about this on [email protected] > recently, specifically regarding the "/site-meta/" or "/host-meta/" > path, which was eventually changed to the more obscure "/.well-known/". > See http://tools.ietf.org/html/draft-nottingham-site-meta for details, > especially versions 00 and 01 of that spec. > > Although it's my impression that BOSH services are not usually served > from the same domain as web services (e.g., once we deploy BOSH at > jabber.org we'll probably serve it from bosh.jabber.org instead of > www.jabber.org), we might want to choose a path that is unlikely to > otherwise show up in the wild. That probably rules out /bosh but > /httpbind might be OK since httpbind is the old shortname for the > technology, as in the http://jabber.org/protocol/httpbind/ namespace. >
If /httpbind then it may as well be /http-bind, I know at least two vendors who won't have to change a thing that way :) Matthew
