[Standards] XEP-0369 (MIX) and sorting out co-existence of MUC and MIX on a single domain

Dave Cridland dave at cridland.net
Wed Jan 11 08:38:30 UTC 2017


On 11 January 2017 at 08:01, Piotr Nosek
<piotr.nosek at erlang-solutions.com> wrote:
> On Tue, Jan 10, 2017 at 3:43 PM, Dave Cridland <dave at cridland.net> wrote:
>>
>> On 10 January 2017 at 14:37, Kevin Smith <kevin.smith at isode.com> wrote:
>> >
>> >
>> > On 10/01/2017 14:27, Dave Cridland wrote:
>> >>
>> >> On 10 January 2017 at 13:30, Kevin Smith <kevin.smith at isode.com> wrote:
>> >>>
>> >>> On 10/01/2017 12:05, Steve Kille wrote:
>> >>>>
>> >>>> I have just issued a PR for MIX version 0.6.4.
>> >>>>
>> >>>> There is clear desire to have the option for  MUC and MIX to use the
>> >>>> same
>> >>>> domain.    The difficulty in achieving this was incompatible disco
>> >>>> results.
>> >>>> This version has made a change to
>> >>>>     add node='mix' to channel disco that will enable the queries to
>> >>>> be
>> >>>> disambiguated.
>> >>>
>> >>> I haven't been able to think of a case other than disco#items on the
>> >>> room
>> >>> JID where MUC and MIX are likely to collide. This change doesn't make
>> >>> it
>> >>> *easy* to implement both on the same domain, but I think it makes it
>> >>> viable
>> >>> - please shout if anyone can think of other cases.
>> >>>
>> >> I agree. Further, I only know of a single client that would ever hit
>> >> disco#items on a room, and that's Psi in its generic disco "browser".
>> >>
>> > Are you suggesting that this approach isn't necessary, and it'd be
>> > sufficient to 'break' disco#items handling for MUC-only clients?
>> >
>>
>> I'd not thought of this approach, but I was considering advocating
>> "just break". I think this means we don't have to.
>
>
> What about using Entity Capabilities to establish whether the client should
> receive MIX or MUC stanzas and syntax? I know that it's mandatory for every
> client to announce its caps but in such case the server could failover to
> default mode. I don't know unfortunately if all major clients include their
> version in initial presence...
>

You don't need to care - a client will either join a MIX using MIX
syntax, or else join a MUC using MUC syntax, hosted at the same
address. I think that with the disco change Steve has made, the two
protocols have no overlap.

The outlier case is a client joining a MUC via GC syntax, but I think
that's practical too (I just haven't thought much about it).

> _______________________________________________
> Standards mailing list
> Info: https://mail.jabber.org/mailman/listinfo/standards
> Unsubscribe: Standards-unsubscribe at xmpp.org
> _______________________________________________
>


More information about the Standards mailing list