On 20.04.2015 17:00, Georg Lukas wrote: > * Florian Schmaus <[email protected]> [2015-04-20 15:27]: >> Contra: >> - Messages and IQs could be used instead >> - Can't be used with BOSH > As you pointed out below, they can be used in theory. I just assume that > most implementations will not expect them and might break in subtle ways. > IIRC, it required significant refactoring of the Smack XMPP library to > accommodate them. I'm sure similar effort will be required in most other > XMPP client and server implementations.
When I added support for SM to Smack I needed to define an internal API for Nonzas anyway, i.e. there is now a "callback" (it's not really a typical callback), which interested parties can consume when a Nonza arrives. So you basically have the effort of adding support for Nonzas in your XMPP stack anyways (as soon as you add support for SM). >> - Introduces a bunch of conceptual and implementations problems > One specific problem is that they can not be accounted for in XEP-0198, > and therefore it is not clear if a Nonza got successfully delivered to > the recipient in case of stream resumption. In the CSI discussion this > caused confusion and led to the notion of resetting the CSI state on > stream resumption, which looks like fixing the symptoms. It appears the author(s) of CSI think of not keeping the CSI state as a feature and specified the protocol that way intentionally, not as workaround for Nonzas not being tracked by SM. >> Pro: >> - Expresses the semantic that they are not routed >> - This increases security, as they are harder to spoof > > I understand the first two points, but I'm not sure if they really > outweigh the problems. I think the advantages outweigh he problems. In fact, I think there are no problems at all. >> A. Nonzas MUST NOT be used after resource binding > With the obvious exception of XEP-0198, of course. Course, SM is fine as it is, And we all ♥ it (I really do). :) - Florian
signature.asc
Description: OpenPGP digital signature
