[Standards] XEP-0115 redux

[Rachel Blackman]

On Jan 16, 2008, at 3:22 AM, Richard Dobson wrote:

>
>>> save that much
>>> bandwidth if we send all the information all the time instead of  
>>> sending it
>>> as an answer to iq:version flood? The flood at least can be aimed  
>>> only at
>>> contacts triggered by the user's interest (like showing tooltip,  
>>> getting user
>>> info, opening the conversation window,...). I actually wouldn't  
>>> call it flood
>>> then, anyway.
>>
>> I used to use that method -- only querying if you were hovering  
>> over the contact and thus a tooltip was needed -- but you can have  
>> a delay in the response, and I ended up with a bug report of 'the  
>> first time I mouse over contacts, their client information doesn't  
>> appeal!'
> But surely in that case then you have the tooltip appear immediately  
> saying "Requesting..." which changes to the client version once its  
> received.

It did, and I still got the bug report from six different people.  I'm  
not necessarily saying this is the wrong way to do it, but sharing a  
data point from actual experience with end-users relating to this  
behavior.

(And if we feel that only querying on-demand when a tooltip needs it  
and sticking to iq:version for that /is/ the right behavior, I'd argue  
we should then document it as a best-practice XEP.  Otherwise, people  
who haven't watched this thread /will/develop clients that do version  
floods because 'that way I get all the data ahead of time' or  
something similar.)

>> In addition, one of the more common usage scenarios seems to be  
>> that various clients use the client information to display a  
>> default avatar if the user doesn't have one set (i.e., showing the  
>> Psi logo for Psi users, an iChat chat-bubble for iChat users, a  
>> Google chat-bubble for Google Talk users, and so on), in which  
>> case, the 'I need client information' usage case is 'the person  
>> came online.'
> In this case you can surely just use the node which is provided in  
> caps anyway?

This is true!  So that particular usage case is still addressed by new  
caps, you're right.

-- 
Rachel Blackman <rcb at ceruleanstudios.com>
Trillian Messenger - http://www.trillianastra.com/