For face image data, there are at least three different issues:
- the underlying format of the face data (i.e., xpm)
- how it is compressed and encoded for transmission (i.e., what goes in
the X-Face header)
- where it is actually stored (X-Face, X-Image, a message annotation...)
and then after all that
- what features the display program has
The first three are appropriate for an RFC (or two). The last is not.
I think it's important to keep all four as distinct as possible.
I think that playing sounds and retrieving faces from a database count as
features of the application. There doesn't need to be much agreement on the
details of that; someone just needs to go an implement it. It can be fixed
later. But when you're talking about data formats that lots of people are
going to be generating, you pretty much need to get it right the first time.
-- Jamie