Re: [xmpp] 3921bis: probe + unavailable

[Peter Saint-Andre <[e-mail]>]

On 2/1/10 6:03 PM, Justin Karneges wrote:
> On Monday 01 February 2010 12:46:52 Peter Saint-Andre wrote:
>> I understand that the MUST here is an imposition on servers. IMHO the
>> deterministic nature of the unavailable has many benefits, but I can see
>> why large service providers would prefer to return a stanza to the
>> probing user only if the contact is online. So I think it's good that we
>> discuss the tradeoffs here.
> 
> A large benefit is to prevent a jabber:iq:last storm.  We don't want there to 
> be any reason for a client to poll its entire roster.

I completely agree with that sentiment!

> I can see why a mobile client might not be interested in having that 
> information pushed at it on login, to reduce resource usage.  Jonas' 
> suggestion of making this a client-controlled option is probably best, but I 
> don't know what the default behavior should be.

Right now we have no protocol for management of account settings such as
this (although we could use ad-hoc commands, XEP-0050). I don't think we
want to introduce such a dependency now, at least not for core delivery
semantics.

> Also, there has been discussion in the past about sending presence only of 
> certain contacts, groups, etc, when mobile, so the solution for filtering out 
> unavailable presence might fit right in with that maybe.

Maybe. :)

> Another thing: this issue is actually two-part: 1) should remote servers reply 
> to probes if the contact is unavailable?  and 2) should the local server send 
> unavailable presence to contacts that just logged in?  Both of these have 
> resource usage implications, but only the second one impacts the client.

I agree with Jonas that these are so closely connected as to be worth
keeping together in the spec.

Peter

-- 
Peter Saint-Andre
https://stpeter.im/