[Freeswitch-users] Problem with subscription expire
Michael Jerris
mike at jerris.com
Sun Oct 4 14:34:01 PDT 2009
This sounds like a bug in the snom to me, we keep changing the expire
on to the future so it should never expire in the first place. You
will have to look at a longer running sip trace to see what exactly is
going on.
Mike
On Oct 1, 2009, at 4:52 AM, Helmut Kuper wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> Hello,
>
> it seems exired subsciptions are never cleared in FS.
>
> A look into sofia_presence.c confirms explains this
>
> /* negative in exptime means keep bumping up sub time to avoid a snafu
> where every device has it's own rules about subscriptions
> that somehow barely resemble the RFC not
> that
> I blame them because the RFC MAY be amibiguous and SHOULD be deleted.
> So to avoid the problem we keep resetting
> the
> expiration date of the subscription so it never expires.
>
> Eybeam completely ignores this option and
> most other subscription-state: directives from rfc3265 and still
> expires.
> Polycom is happy to keep upping the
> subscription expiry back to the original time on each new notify.
> The rest ... who knows...?
>
> */
>
> For some reasons subscriptions created by Snom phones are filling up
> the
> sip_subscriptions table over time. This leads to some kind of DOS by
> FS
> against the subscribing phone ... The subscribtions are
> differentiate by
> call-id. This can be explained by RFC 3842 chapter 3.6 where expired
> subscriptions must be renewed with a NEW call-id. Because there is no
> hint about unsubscribing the old subscription I guess the clean up
> process has to be done by FS.
>
> Any way to get FS to do this job? Since there is no creation date or
> expire value which represents the expire as a timestamp I have no
> way to
> clean up the table manually via sql and cronjob - except cleaning the
> whole table ...
>
>
> A further (but background) question is, why do the subscriptions
> expire
> in snom phones at all ...
More information about the FreeSWITCH-users
mailing list