Matthew Wild wrote: > > Your earlier mail about stream features indicated that no child > element of a feature would imply the feature is optional and not > mandatory for further session negotiation. The RFC now seems to say > the opposite ("(i.e., features default to required)", line 517). Which > is intended to be the case? or did I miss something?
We had a chat about this today in Psi discussion room: http://conference.psi-im.org/[EMAIL PROTECTED]/2008/10/16.html#13:37:26 The problem is that current usage is all over the map. As far as I can see, the existing stream features have the following implicit defaults: STARTTLS = optional SASL = required resource binding = required IM session = required stream compression = optional The defaults matter for all existing software because it will take a long time for software to be upgraded for compliance with rfc3920bis. No matter what we do, we'll need to "grandfather" some features by exempting them from the default handling rules. If we say that stream features are optional by default, then we need to exempt SASL, resource binding, and IM session establishment. If we say that stream features are required by default, then we need to exempt STARTTLS and compression. Pick your poison. :) Peter -- Peter Saint-Andre https://stpeter.im/