[Standards] NEW: XEP-0209 (Metacontacts)

Etan S. C. Reisner deryni at unreliablesource.net
Wed Jun 6 08:22:07 CDT 2007


On Wed, Jun 06, 2007 at 10:47:38AM +0100, Kevin Smith wrote:
> Somewhat delayed reaction.
>
> On 5 May 2007, at 10:21, Etan S. C. Reisner wrote:
> >Perhaps I've just missed it but how does this interact with roster
> >groups?
>
> The intention was that clients would be able to render the group
> membership of the metacontact in some appropriate way of their
> choosing. I'm happy to add an implementation note if people think
> it'll clarify matters.

I personally wouldn't want my contacts being moved between groups if I had
placed them in separate groups and so would like for metacontacts to be
usable per-group. Further comment on this in my next point.

> >Is the intention that meta-contacts are defined across all groups?
> >So that
> >if two jids are set as a meta-contact they should be placed
> >together in
> >all groups that they share? Or do meta-contacts span groups such
> >that jids
> >from multiple groups are pulled together into one meta-contact in
> >one of
> >the groups? (Which seems like it would be confusing to me so I'm
> >assuming
> >was not the intention.)
>
> I imagine the most used implementation would be to show the
> metacontact in the group(s) of the primary contact, but any other
> rendering is equally valid, to my mind.

Wouldn't that interpretation of the XEP basically forbid (or at least make
non-deterministic) the showing of a buddy in multiple roster groups?
Assuming a buddy is both in a metacontact (as a bare jid) and in multiple
roster groups, which of the buddy list copies is to be combined into the
metacontact? The first one we get from the server? The 'top' one in the
roster group ordering? All of them?

> >Also, given that a jid can be placed in multiple groups at once
> >what is
> >the reasoning behind "any jid MUST NOT be specified in this manner
> >as a
> >member of more than one metacontact within an account"? Was the
> >intention
> >specifically to forbid having jids in different metacontacts in
> >different
> >groups (assuming that metacontacts don't span groups)? Or was this
> >just to
> >keep implementations simple?
>
> The intention was unrelated to group membership; it was simply that a
> contact should only belong to one metacontact; since these
> metacontacts are a method of pulling the different representations of
> a single entity together, having one jid in multiple metacontacts
> doesn't really make much sense.

While I mostly agree that it doesn't make much sense I can imagine having
a buddy with multiple home and work accounts and wanting different
groupings of them in my 'Home' and 'Work' roster groups. This is not at
all a likely scenario but I can imagine someone wanting it.

I'm less concerned with this restriction than I was when I first read it
(especially given that I don't remember the exact concern I had with it at
the time =) so while I still am not sure this is needed I no longer have
objections to it.

> /K

	-Etan


More information about the Standards mailing list