[Standards] Proposed XMPP Extension: MUC Auto-Join
ian.paterson at clientside.co.uk
Mon Jun 4 16:38:45 UTC 2007
Do we really need presence-triggered invitations?
The arguments expressed here seem to be:
1. Room presence publishes the room capabilities
2. The user knows whether room is online or not
3. The auto-invite works with existing clients - and will therefore be
deployed more quickly
IMHO none of these seem to be killer arguments:
1. I agree with other people here that the identity information (the
fact that an item is in fact a room not a user) should be available even
if the room is offline. Presence-based caps doesn't give us that. Also,
IMHO, at least some of the extra functionality that Chris is talking
about is generally useful and could be usefully standardized. That
probably means storing room/user specific parameters server-side.
(Server-side storage is a good thing because it is available to all the
clients you use.) As Peter intimated, allowing rosters to be extended in
any way is probably too big a change for us to include in RFC 3921bis.
So I'm in favor of a new fully-featured "Room Bookmarks" XEP (including
optioanl auto-open and much more).
2. Does this matter? Are MUC servers really that unreliable that the
user will really care if she (occasionally) only discovers that the room
is offline after her client fails to enter it automatically for her?
Admitedly, presence invite does enable a "join as soon as the room comes
back up" feature. But if room servers are so unreliable, perhaps we're
not addressing the real problem here? And caps is certainly not the way
to discover if the JID is a room or a user if the room is likely to be
3. But it doesn't work for any existing MUC servers. During the phase
before widespread adoption of whichever protocol we decide on, if a
(power) user wants to take advantage of auto-enter functionality then it
will far easier for her to change her client than to change her MUC
server. In fact, she will have no control over which server hosts some
of the rooms she visits. (For example, regular jdev visitors would be at
the mercy of the XSF until conference.jabber.org is upgraded... and XSF
would probably wait at least until the ejabberd team implement it.)
While I'm not strongly against presence-triggered-invites, and am open
to be convinced, there seem to be some drawbacks. IMHO there are better
ways to give users (fully-featured) auto-enter.
More information about the Standards