RE: Generic proposal for enganging MTOM in WSDL 2.0

Hi Jonathan!

> [this] goes against the style encouraged by WS-Policy specs, which is to use
> PolicyReference elements pointing to top-level policies.  This is
> illustrated in the primer [1].  I think this style is more readable and
> maintainable than embedding policy expressions inside WSDL operations, and
> seems to be the current practice on the ground.  

The flag isn't intended to be applied to any old WS-Policy rather one
which a publisher decides to craft in a way that's digestable 
by a non-WS-Policy processor.

> The profile of policy that
> the proposal below implies doesn't match this style, and therefore it's
> unlikely to be as broadly interoperable as we'd like. 

Is the risk that WS-Policy processors are unlikely to 
support the simple inline WS-Policy style?
 
Hi Arthur!

> [snip] it seems to me that you are proposing
> to profile WS-Policy. 

Oooh "profile" is such a loaded word.. I'm not saying to the world
"don't use WS-Policy", I'm saying to Canon "don't stick your MTOM assertions
firectly into WSDL, wrap them in a WS-Policy element and you'll interoperate
with WS-Policy processors" ..

> Wouldn't it be better if the WS-Policy WG defined a
> simple subset so that simple processors could implement it? This is like SVG
> Tiny. Maybe we need a WS-Policy Tiny? 

That would be one approach, but this isn't really for someone 
who is WS-Policy aware .. the flag could be applied to *any* wrapper
element, really. WS-Policy is a for-instance, we don't have to tie it down 
to one particular wrapper element QName. 

The aim of the proposal is to allow a publisher to write WSDL 2.0s which 
they can interoperate with WS-Policy aware processors, but allow them to indicate
to a consumer that they don't need to understand WS-Policy
language, that it's safe to just look for the precense of one or 
two XPaths to see if MTOM, etc are engaged.

Paul

Received on Wednesday, 29 November 2006 14:39:58 UTC