Derek, There is still a generic problem for a dialog event package event state compositor that puts together <dialog> elements from multiple UAs that the id attributes may not be unique. This id is used for partial dialog updates, so it must be unique across all dialogs for an AOR, which is not what RFC 4235 says. However, the specific problem relating to the new <joined-dialog> element isn't a problem any more.
I wonder if this is a bug in RFC 4235 or whether it is just a problem that event state compositors have to deal with. Thanks, Alan Derek MacDonald wrote: > There is still some text in 6.3 which mentions dialog id conflicts; is > that just stray text? > > -Derek > > On Thu, Feb 28, 2008 at 3:47 PM, Alan Johnston <[EMAIL PROTECTED]> wrote: > >> Derek, >> >> It is already in the draft that way - I thought of this when I did the >> schema. >> >> Thanks, >> Alan >> >> >> >> Derek MacDonald wrote: >> > Section 6.1.3 describes a <joined-dialog> element which uses the id >> > attribute of the dialog element to identity which dialogs are joined. >> > Unfortunately, the id attribute is only unique for a UA instance. This >> > requires that the state compositor MUST be prepared to offer different >> > notification documents(for the same dialog) for the UAs that have >> > conflicting id attributes. >> > >> > An alternate approach would be to not use the id portion of the dialog >> > element to identity a joined dialog. It would still have call-id, >> > local-tag, remote-tag and direction. This eliminates the need for a >> > state compositor to provide different views; in fact, no composition >> > will be required within a dialog-info element. >> > >> > I am not sure why we didn't do that in the first place. >> > Interestingly, I couldn't find any text in rfc4235 around id attribute >> > collisions. It seems they lose meaning when different UAs sharing an >> > aor are composed in a central state agent. >> > >> > -Derek >> > _______________________________________________ >> > BLISS mailing list >> > [email protected] >> > https://www.ietf.org/mailman/listinfo/bliss >> > >> > >> > >> >> >> > > > _______________________________________________ BLISS mailing list [email protected] https://www.ietf.org/mailman/listinfo/bliss
