[Telepathy] Avatar interface proposal

[Raphael Slinckx]

On Thu, 2006-06-08 at 09:40 -0300, Andre Magalhaes wrote:
> Hi,
> 
> On 6/7/06, Rob Taylor <robtaylor at floopily.org> wrote:
> > I would say:
> > GetAvatarId (u: contact ) -> s: avatar_id
> >
> > The avatar id should be a unique, or statistically nearly unique
> > identifier for a given image. It should also be valid over multiple
> > sessions. e.g. a hash of the image data.
[snip]

While i think the proposed API with the changes is reasonable, don't you
think this is a major abstraction leakage ?

I know that for practical purposes (imposed by jabber's supposedly
broken implementation only) it is easier to do it this way (having the
client do the dirty work).

Still, i think the telepathy API shouldn't reflect that problem, and
that the 'workaround' should be implemented in the jabber connection
manager, to be one day removed when jabber gets fixed (read: never).

That implies that the jabber connection manager has a state containing
old avatars and that it can do the matching against new hashes and all,
to finally emit a new avatar signal.

Also, does anyone have any insight about how other protocols work, do
they make the same mistake as jabber, or do they work correctly ?
Maybe it will be actually more difficult to bend correctly-behaving
protocols to the twisted xmpp behavior leaked in the spec ?

Raf