[Standards-JIG] Re: [jdev] XMPP Ping method?

Mridul mridul at sun.com
Thu Nov 2 15:11:28 UTC 2006


Kevin Smith wrote:
> On 2 Nov 2006, at 12:49, Mridul wrote:
>> For reliable delivery - you could always look at end to end 
>> acknowledgment.
>
> I've said before, and I think it needs to be said again: e2e 
> acknowledgement provides accountability, not reliability. 

If you have accountability - you also buy in reliability.
Though reverse is not true ofcourse.

> Reliability just means you know that the message will be delivered. 
> Telling the remote entity when you receive a message opens huge issues 
> for non-subscribed users - either you get presence leaks large enough 
> to drain the atlantic, or you lose reliability. 

I would consider e2e ack to be something which does not come 
'automatically' - but something which gets negotiated by both entities 
explictly.
Else you will not be able to interoperate with servers/clients which do 
not support it.
While negotiating it , you could decide to enable/allow/disallow it from 
'untrusted' entities - so there need not be a leak.

> I'm completely convinced that per-hop is the way to go.

per-hop has the issue that over a multi hop delivery, it might be 
necessary for the sender to know of the delivery failure.
AMP could be used to alleviate this problem to some extent , but I am 
not sure if that would work for every use case.
To propogate a per-hop delivery failure back for all usecases might be 
slightly complex imo ...

Regards,
Mridul

>
> /K
>
> --Kevin Smith
> Psi XMPP Client Project Leader (http://psi-im.org)
>
>
>




More information about the Standards mailing list