CR33: outline of a proposal for casting wsaw:Anonymous in more policy friendly terms


Paco and I took an AI last week to take a shot at casting the 
wsaw:Anonymous WSDL marker in more policy friendly terms. We have agreed 
on most of the outline for a proposal except for two points that are 
called out below. We don't have a detailed proposal with changes to the 
existing text yet since there are some details to be worked out, but it 
would be fruitful in having a discussion on whether this direction makes 
sense to folks.

1) Replace the wsaw:Anonymous element with multiple WSDL markers that 
can also act as policy assertions a la wsaw:UsingAddressing.

2) The markers/assertions would *not* be coupled to wsaw:UsingAddressing 

3) The markers/assertions would specify the ability to send a response 
on the 'backchannel' and are not tied to the WS-A 'anon' uri. This part 
is similar to the requirement relaxation made in the 1st proposal [1] 
sent by Doug and me. The marker would be about the ability to open new
connection v. sending stuff back in the 'backchannel' (ws-a 'anon' uri 
being one way of specifying that).

Discussion points:

1) Should there be three assertions/markers or two?
The question is around what happens when none of the markers are specified?
One opinion is that there should be two markers that say: backchannel is 
required OR a new connection is required. An absence of a 
marker/assertion (default) says that both are supported. This way of 
specifying markers has its rationale in the current thinking in the 
WS-Policy WG which says that absence of an assertion is negation. This 
thinking is also based in the view that these markers are restricting in 
what is allowed by ws-addr and therefore absence of the marker/assertion 
means everything is supported/allowed.
Another opinion on this is that there should be three markers 
corresponding to the 'required', 'prohibited' and 'optional' values of 
wsaw:Anonymous and default value should be unknown. This thinking is 
based on the fact that the new markers/assertions are decoupled from 
wsaw:UsingAddressing and disagreement with 'absence is negation' way of 
thinking about assertions/markers.

2) What should be the name of the markers/assertions?
Two proposals are:
a) wsaw:ResponseOverNewConnection, wsaw:ResponseOverBackChannel, 
b) wsaw:NewConnectionRequiredForResponses, 




Received on Monday, 2 October 2006 18:45:20 UTC