[jdev] XEP-0199 (XMPP Ping) to test reliability

[Yann Leboulanger]

Dave Cridland wrote:
> On Mon Jan 21 18:52:54 2008, Yann Leboulanger wrote:
>> In gajim we send whitespace ping if we haven't received or sent 
>> anything in the past 55 seconds (cause some nat server close 
>> connection if nothing happen in a minute)
>> But whitespace ping are not enough, so replacing it with xmpp-ping 
>> with the same time would be nice.
>>
>>
> Mhh. Okay. Whitespace pings aren't enough to tell if the connection is 
> actively able to send and receive packets. XEP-0199 tells you not only 
> that, but it also tells you whether the thing you're pinging is willing 
> and able to respond.
> 
> Both have a use, although for c2s links, XEP-0198 is rather more powerful.
> 
> Don't confuse those use-cases, because whether or not you use XEP-0199 
> to test c2s connectivity, whitespace pings are still lighter, and 
> perfect for keeping recalcitrant NATs in line.
> 
> XEP-0199 is particularly useful when you're expecting a response, but 
> don't seem to be getting anything.
> 
> 
>> about the time for answer, are some network connection or server so 
>> slow that it can reply only 20 seconds later? I have no feedback on 
>> that, but isn't 5 or 10 seconds enough?
> 
> HF radio links would need much more, whereas a DSL link would need less.
> 
> A good rule of thumb might be 10 times the normal RTT. (Which you can 
> detirmine by the usual response to XEP-0199 pings).
> 
> IMHO, a nice UI would simply note that the latency seemed tremendously 
> high, and offer to reconnect, rather than kill the session - as Michal 
> pointed out, the user often knows what the situation is.
> 
> Dave.

Ha yes I didn't know XEP 198 has a ping section. But we don't get an 
answer, at least with my ejabberd. And in this case an answer is 
usefulle to know if server is still alive. So I think XEP 199 is better.

-- 
Yann