RE: More thoughts on making binding detail definition reusable

Hi Jacek,

Thanks for the feedback. The revised proposal was intended to facilitate discussions since the original proposal was listed in the agenda for last week's F2F. Reasonably that discussion was postponed again due to dependancies on our decisions in related pending binding issues. Once the WG reaches
resolutions on the binding component model, the syntax proposed here should be adjusted accordingly.

I am open to the idea of allowing a binding and its child elements to reference multiple bindingTBD elements. One problem I see is potential conflicts between refrenced data, for example both referenced bindingTBD specify a value for "style", one use "rpc", one use "document". We need to clearly
state the rule of resolving such conflicts.  

As for the optionality of the operation name attribute, thanks for pointing it out. I just copied it from Sanjiva's original servicetype based proposal. It should be changed to be inline with our final binding component model. 

Best Regards,
Kevin



-----Original Message-----
From: Jacek Kopecky [mailto:jacek@systinet.com] 
Sent: Tuesday, January 28, 2003 4:12 AM
To: Liu, Kevin
Cc: 'www-ws-desc@w3.org'
Subject: Re: More thoughts on making binding detail definition reusable


Kevin,

the original Sanjiva's proposal that started the discussions (IIRC) [1]
also allowed specifying multiple bindingTBDs (in your terminology) to be
used in a concrete binding. I think the reasons for this were good, and
for example Jean-Jacques's favourite split of serialization and protocol
bindings would be doable (a binding referencing a serialization
bindingTBD and a protocol bindingTBD).

So in terms of syntax, the change to your proposal would look like

<binding name="ncname" type = "qname" TBD = "list of qnames">* 
  <operation name="ncname"? TBDs = "list of qnames">*
    <input TBDs = "list of qnames"/>
    <output TBDs = "list of qnames"/>
    <fault name="qname"? TBDs = "list of qnames">
  </operation>
</binding>

i.e. changing the new attribute's value to a list of qnames.

Also, I noticed that operation name is optional in your structure - have
I missed something? I think the name is mandatory.

Best regards,

                   Jacek Kopecky

                   Senior Architect, Systinet Corporation
                   http://www.systinet.com/


[1] http://lists.w3.org/Archives/Public/www-ws-desc/2002Jul/att-0117/01-bindings-2002-07-24.html

Received on Wednesday, 29 January 2003 17:23:41 UTC