allowing me to execute SASL negotiations fully asynchronously within djabberd. Tests are still expecting a linear run
to return 1 for PLAIN authentication, otherwise this is sent to the client which confuses some implementations.
to change qop for the server
takes a hash with no_integrity and no_confidentiality configurations.
I'm still not sure about this change. I'm also pondering about renaming set_success to something less ambiguous. We might need better semantic in the client, where success of a negotiation is communicated to the client thru specific protocol means sometimes.
the system doesn't support it and vice-versa
It has to be noted that getsecret is a deprecated callback in SASL v2 (apparently).
is proposed only, but that should be fine) - Fixed an issue (I believe) where the client was looking for 'server-maxbuf' insted of 'maxbuf' - renamed a test (consistency)
client and one Authen::SASL server - Fixed realm challenge parsing when the challenge originates from the client. In this case it shouldn't be a multivalued token
is consistent with the Cyrus implementation.
the EXTERNAL mechanism has only one round trip so that the authorize-id needs to be send in client_start. client_step will never get called and can be removed. Also, the noanonymous flag can be set. See attached patch. Signed-off-by: Graham Barr <email@example.com>