[standards-jig] NEW: Jabber Object Access Protocol (JEP-0075)

evan at prodromou.san-francisco.ca.us evan at prodromou.san-francisco.ca.us
Fri Mar 7 10:16:04 UTC 2003

On Wed, Mar 05, 2003 at 10:16:05PM -0600, Peter Saint-Andre wrote:

> Evan Prodromou has contributed a big JEP that defines methods for creating
> generic object-access servers in Jabber. Fascinating reading. Who needs
> SOAP when you've got JOAP? ;)
> http://www.jabber.org/jeps/jep-0075.html

Hooray! I'm glad to see this up on the site. I hope it stimulates some
comment -- I'm interested to hear how they go.

I have some particular questions that I'd like to draw attention to, some of
which are in the "Future Considerations" section, and some I hadn't added yet:

1) Is the XML-RPC basis of JOAP appropriate? I find XML-RPC to be a
beautiful specification, and it inspired most of JOAP (doing just enough to
be complete, without making things too complicated). However, now that
there's a SOAP JEP, would it be more reasonable to use that type system, and
method format, than XML-RPC's?

2) I mentioned the use of presence to make "live" objects -- letting clients
"subscribe to the presence" of an object, and having the object send updates
to the client when the object state changes. I'm thinking this is a great
advantage XMPP has over an HTTP-based object system, since the objects can
actually "talk back" to the client. I think it'd be great for, for example,
monitoring, or caching. But I'm also wary of needlessly complicating the spec.

3) The introspection mechanism -- <describe> -- is kind of terse about
structs and arrays. I think that there might be some value in describing the
members of a struct or array.

4) There's no provision for locking of objects. Concurrency is a real bag of
worms when you're dealing with a distributed data system, and I'm torn
between adding a <lock>-<unlock> pair of verbs, or just letting it float.

OK, that's the main stuff on my mind right now. I'm all giddy.


More information about the Standards mailing list