On 20 April 2015 at 17:39, Georg Lukas <[email protected]> wrote: > * Matthew Wild <[email protected]> [2015-04-20 18:33]: >> > In the XEP-0352 case, "Nonzas" could get lost without resending them >> > upon stream resumption, leaving the client under the impression that >> > it's active, although the server thinks it's inactive. >> This can't happen, because new streams are always active by default. I >> really thought this was clear in the XEP, but actually it's a little >> bit ambiguous. > > This looks to me like a restriction caused by using nonzas in the first > place (and the inability to track them in 0198), not like something > which makes sense from a protocol desgin view point. > > As a client developer, I would expect the stream to be in the same state > it was before resumption, as stream resumption is caused by network > interruptions and has nothing to do with user/client activity.
It depends whether you view CSI as a property of the stream or not (which it is currently considered to be). If you "expect the stream to be in the same state", would you also expect to not have to renegotiate compression on a resumed stream? Regards, Matthe
