[Standards-JIG] avatars

Richard Dobson richard at dobson-i.net
Fri Jul 2 11:49:04 UTC 2004


> I see, that I did omit the hashes. Hashes are important and we use them
heavily
> exactly for what you say. My client fetches avatars from many users. This
is done
> only if the hash does not match. (These <obj/>s were just example what I
mean
> by harmonization. I did not propose the next protocol. They were for sure
neither
> complete   nor correct)

Ok, good, yup the filehashes must stay.

> Also there are obviously 2 notions of in-band here. I agree that data
objects
> should not be in-band when something (XHTML) is pushed. In this case there
> should be the hash. Which is not there. I only see cid and id in
>  <obj xmlns='http://jabber.org/protocol/iobj'/>. But if I decide to
retrieve a data
> object after  disco, checking for size, mime-type, etc. If I decide i want
the data
> object then I would like the idea to be able to have the data-object
in-band in the
> response. Please note that I am NOT pushing for in-band only. I am still
trying to
> get systematics into the way to request data objects in different Jabber
sub-protocols.

If you want to receive the object inband then you just follow it along and
at the file transfer stage just specify that you want it using IBB, IMO it
should be kept nice and layered like it is now re-using existing protocols.

> Yes in case of XHTML the author could simply put an HTTP URL. But this
only
> solves the problem for XHTML. Still avatars can not be fetched via HTTP.
Avatars
> over sipub is OK from client to client, but avatar galleries would work
well with HTTP.
> So how do you specify the HTTP URL for an avatar? And I repeat: avatars
are just
> the example. Soon we will design YAJP (yet another Jabber protocol), which
retrieves
> data objects via sipub, in-band or OOB. There will be a YAJP-Wiki and
peopel will
> discuss again the same issues as for avatars.

Then maybe we need to extend obj slightly for more general use for HTTP
URL's as follows

<obj xmlns='http://jabber.org/protocol/iobj'
id='552da749930852c69ae5d2141d3766b1' src='http://jabber.org/image.gif'/>

> So, there really should NOT be an avatar protocol. There should be a more
general one.
> A more general protocol means that it should support different methods in
a harmonized
> way. It will probably be just a wrapper for IBB, sipub, or XHTML-obj, and
some disco.
> Maybe some have to be adapted, maybe not. But if there is an avatar
protocol, then we
> will get YAJP soon after, which will be very similar.

If you read my previous email you will see that there is no need to invent a
new generic wrapper as IMO our XHTML obj spec already provides a pretty
generic framework, it is fine as a wrapper for file transfer (no need to be
a wrapper for IBB as by using filetransfer it will automatically pick the
appropriate transfer method be that OOB or IBB).

Richard





More information about the Standards mailing list