[Standards] XEP-0198 status

Evgeniy Khramtsov xramtsov at gmail.com
Thu Jan 13 05:51:09 UTC 2011


13.01.2011 05:26, Peter Saint-Andre wrote:
> That seems mostly reasonable to me. I see a possible edge case if the
> reconnecion address (ip:port or host:port or whatever) isn't actually
> there by the time the initiating entity tries to reconnect, i.e., the
> server tells me the reconnection address and a few days later when my
> session dies I finally try to reconnect but the address that the server
> *thought* was going to work a few days ago isn't so much alive anymore.
> Do we need to worry about that? If so, does the initiating entity just
> follow the rules from 3920bis regarding the connection process? (See for
> instance the rules for<see-other-host/>  in 3920bis...)

I think we need more formal description on this. An *informational* XEP 
about load-balancing and clustering would be nice.
The XEP should describe:
1) possible ways of making clustering and load balancing, describing 
problems with sharing XMPP-specific data, so server implementors has a 
variety of options and a complete picture.
2) what technics are preferable (<see-other-host/> from the RFC, 
proposed <enabled/> from XEP-0198, redirect technics in BOSH and so on) 
in order to make sure clients can correctly handle this.

I can prepare initial raw version of the XEP if someone of you guys 
wishes to participate in this :)

-- 
Regards,
Evgeniy Khramtsov, ProcessOne.
xmpp:xram at jabber.ru.




More information about the Standards mailing list