I've not looked at the specific proposal yet, but my skepticism
persists.
Just because we're unable to to test a proposal to mandate something,
we've gone down the path of:
- Mandating support for WS-Addressing
- Mandating support for EPR where we used to just suggest it.
- Mandating support for the WS-Policy indication that flags the use
of WS-Addressing
- Mandating support of the protocol assertion for the protocol when
support is there
... all around an issue where we all seem to agree that the use cases
are unclear. My design instincts are screaming "feature creep!" All
of this nets out to an implementation needing to recognize the
WS-Addressing assertion in a concrete referenced WSDL, and then using
the support that we've now mandated. It doesn't actually reveal much
about actual support for the underlying concern - WS-Policy. The above
set of mandates does reveal the ability to recognize XML elements in a
particular scenario and not barf them up, but that's about it.
WS-Policy is a particular XML-based expression of a model for policies
- a "platform dependent model" (PDM) in UML terms. SCA intents come
close to being a "platform independent model" for policy requirements
as I've seen.
Without a mandate to use WS-Policy, implementers can happily punt on
correlating between the two, and hopefully avoid complexity for
themselves and their customers by always generating one (the PDM) from
the other (PIM). In fact, the way to generate the PDM from the PIM is
to define the mechanism that does the one-way translation. In
mandating WS-Policy, we might make it necessary for implementations to
think about having a bi-directional model between the two, where (a) it
might not make sense, and (b) it may actually be more confusing to the
end-user than simply giving an implementation the freedom to say "I
don't understand how to do what you're asking me."
Does anyone actually have implementation experience that suggests that
this particular mandate works? If so, I will happily hear the details
and how they work, and be quiet. Otherwise, I think we're going to far
with 126.
-Eric
On 05/05/2010 01:09 PM, Anish Karmarkar wrote:
4BE1D06E.6060003@oracle.com" type="cite">Attached.
The proposal uses cd03-rev4 as the basis (with changes accepted). The
relevant changes are confined to section 2.10 (new section) and section
6.4. Do note that Mike & I had taken a joint AI to produce a
complete proposal for issue 126. The attached doc adds support for
ws-addr but not for ws-policy.
I have made one change that was not discussed on previous calls or on
the ML: when the callback protocol is supported I had made changes that
require the runtime to support the callback protocol policy assertion.
Since this proposal is about requiring ws-policy, I thought it made a
lot of sense to mandate support for the protocol assertion when the
protocol is supported.
If this (or something like this) is accepted, I think we should make
the endpointReference element mandatory (currently it is a SHOULD).
Especially, since the UPA issue resolution means that it would be the
same element defined in ws-address. But on the last call, someone
expressed preference to deal with this separately. I'll raise an issue
related to that if/when 126 is resolved.
Mike: I know this proposal doesn't give you a lot of time to add
ws-policy support before the bindings call. Please let me know if you
don't have time and I can try and add that later this evening/tonight
(my time).
Thanks.
-Anish
--
---------------------------------------------------------------------
To unsubscribe from this mail list, you must leave the OASIS TC that
generates this mail. Follow this link to all your TCs in OASIS at:
https://www.oasis-open.org/apps/org/workgroup/portal/my_workgroups.php
|