Thanks for the info Raymond.

...snip
> It seems that WS-AT only cares if the TX context is flowed inside the WS 
> message. I'm not sure how we define the policySets for the implementation
> intents such as sca:managedTransaction.global.

Not sure about this. It says "For application messages that use a SOAP
binding, the Atomic Transaction coordination context MUST flow as a
SOAP header in the message"

Does this mean that

for applications that don't use the SOAP binding we don't care what happens

OR

for applications that don't use the SOAP binding something is expected
to happen but it's not defined by the spec.

I prefer the latter. It's certainly the case that our policy must have
a binding specific flavour. Hence it maybe that we

1/ don't map wsat:ATAssertion to any particular Tuscany policy
provider and rely on binding implementations to know how to deal with
it
2/ map wsat:ATAssertion to a single Tuscany policy provider and have
that policy provider do different things depending on what binding is
in place
3/ map wsat:ATAssertion to multiple Tuscany policy providers, one for
each binding. We can't do this at the moment
4/ Use different policy assertions for the application of  the atomic
transaction assertion to different bindings. In this case
wsat:ATAssertion is reserved for the web service binding

I think we are heading in the direction of 4/ although the code from
1.x, and hence what we need to review in 2.x, does quite a bit of 1/
at the moment.

I don't mind relying on the binding code to consume assertions per se
but it worries my that we don't have a good way of advertising to the
Tuscany user that they are available. This may just be a matter of
documentation but if it came to it that we need policy providers that
did actually do very much then I'd rather have that than nothing at
all.

Simon

Reply via email to