[Standards-JIG] Re: JEP-126 (Invisibility)

Hernan Tylim htylim at novamens.com
Wed Apr 12 18:40:22 UTC 2006


I am not a server implementor, and I know that I might be talking 
without a solid knowledge, but my opinion is that the current Presence 
mechanism is too rigid. To me direct presences shouldn't exist, or in 
any case, they should be handled by the user's server and not just 
routed and ignored.

What I would like, or more precisely need in a near future, is to be 
able to be online to only a group of contacts, busy to others and away 
to everybody else.

I know that I can do this with a combination of direct-presences and 
privacy list, but, as someone already stated on this thread, they are 
awkwards to use for these kind of scenarios.

My proposal is to use IQs to selectively change a user's presence to 
invisible(offline) but also to any other state with an option to choose 
to which user's contacts this presence is directed. Note that if no 
contacts are explicitly choosen then the presence can be globally 

P.S: I think that these scenarios will be more important as Jabber is 
used each time more and more in corporations.


Ralph Meijer wrote:

>On Wed, Apr 12, 2006 at 07:23:02PM +0200, Remko Troncon wrote:
>>On 12 Apr 2006, at 17:23, Peter Saint-Andre wrote:
>>>The traditional approach to invisibility has been to stop presence
>>>broadcasts at my server, so they would never get to your server (let
>>>alone your client).
>>Exactly, that's the whole point of being invisible: that your  
>>contacts don't know you are online. Rather useless if the presence  
>>still reaches them :-) This is exactly why defining it as a special  
>>type is not a clean way to go at it, since this means that the  
>>semantics of presence changes (i.e. presence packets are no longer  
>>broadcasted to everybody), complicating the spec.
>>It's not that because invisibility is not a presence type at the  
>>protocol level, that clients cannot represent it as a special type of  
>>presence. Just leave the 'appear offline' in the UI, and translate it  
>>into <iq invisible>. If the user goes online again, do an <iq  
>>visible> again and send the selected presence.
>I'll assume here that the reasons for wanting invisibility are sound, as
>it seems we may never get a compromise. I'll just comment on the
>technical aspects.
>I am with those that say invisibility is not really a presence setting
>and we should use iq stanzas. So, you still send your presence to the
>local server as usual and use the iq's to steer the broadcasting. That
>is practical and conceptually valid.
>Not that this doesn't invalidate the use of privacy lists. You can still
>block outgoing (presence) traffic, even though it will not be sent when
>going invisible.
>Really, it is much like node configuration in pubsub. I imagine that it
>is also easier to become invisible for subsets of your roster by
>tweaking the protocol.
>Directed presence will not be affected as it is not destined for your
>local server.
>I would choose having this in a JEP rather than the RFC.

More information about the Standards mailing list