Re: Proposal for Describing Web Services that Refer to Other Web Services: R085

Arthur,

Thanks a lot for the comments.

>If you feel that dynamic discovery is important, then I suggest you submit this to the WG as a new >requirement, and give a good motivating example
Well, this is where the hard part begins, it's much easier to suggest various ways if improving your proposal :-), but I'll try to give it a try (sorry for the pun !). 


>IMHO, I think it is useful to dynamically retrieve the WSDL associated with an endpoint, not just retrieve the >binding.
Do you mean that a retrieved WSDL also contains the <service> element ? One issue is that case is that it's not clear what the endpoint address will be, the address which was used to retrieve that WSDL (created on the base of [1], without an extra parameter ?WSDL or smth, if used) or a port address found within the <service> element of the dynamically retrieved WSDL. Probably, the latter one, in which case the meaning of the former one changes at runtime : if <service> element is available then it's just an address from where a WSDL doc can be retrieved, otherwise it is the endpoint address, as defined by [1].

I thought that a simple binding discovery could be dependent on your proposal, that is an endpoint is completely described as shown in [1] anyway, so that a client runtime, which can not, chooses not or fails to retrieve bindings dynamically, can always bind using a statically available info. In that case, @binding="dynamic" will serve as a hint. Again, as you said, a good example would be useful.  

>I see a small problem with your proposed implementation. You say that the binding is obtained by performing >a retrieval on the URI of the endpoint. However, you can't retrieve the URI unless you know the binding. You >have to assume some binding, e.g. if the URI uses http: then do a GET (possibly with something appended to >the URI, e.g. ?WSDL).
If I understand you correctly, then depending on the URI scheme used (the one used for accessing the referencing service), an appropriate query verb/construct should be used for a retrieval; but I'm not sure I do realize now what I'm trying to say :-)

Cheers
Sergey Beryozkin

[1]
http://lists.w3.org/Archives/Public/www-ws-desc/2003Apr/att-0088/R085-2003-04-22.html
  ----- Original Message ----- 
  From: Arthur Ryman 
  To: Sergey Beryozkin 
  Cc: www-ws-desc@w3.org ; www-ws-desc-request@w3.org 
  Sent: Monday, May 05, 2003 6:44 PM
  Subject: Re: Proposal for Describing Web Services that Refer to Other Web Services: R085



  Sergey, 

  You are defining a dynamic binding discovery protocol, which is  beyond the scope of R085. Recall that R085 is a WSDL requirement. It states that a WSDL document must describe the endpoint. If you feel that dynamic discovery is important, then I suggest you submit this to the WG as a new requirement, and give a good motivating example. IMHO, I think it is useful to dynamically retrieve the WSDL associated with an endpoint, not just retrieve the binding. 

  I see a small problem with your proposed implementation. You say that the binding is obtained by performing a retrieval on the URI of the endpoint. However, you can't retrieve the URI unless you know the binding. You have to assume some binding, e.g. if the URI uses http: then do a GET (possibly with something appended to the URI, e.g. ?WSDL).

  Arthur Ryman 


       "Sergey Beryozkin" <sberyozkin@zandar.com> 
        Sent by: www-ws-desc-request@w3.org 
        05/01/2003 11:48 AM 

               
                To:        Arthur Ryman/Toronto/IBM@IBMCA 
                cc:        <www-ws-desc@w3.org> 
                Subject:        Re: Proposal for Describing Web Services that Refer to Other Web Services: R085 

                



  Hello,

  I'd just like to return again to the question of whether dynamic bindings
  should be disallowed by the proposal[1]or not. Proposal [1] says @binding
  attribute is declared statically in the WSDL document, this probably covers
  the majority of cases.

  How practical/usefult would it be to add @binding optional attribute to the
  endpoint reference definition :

  <wsdl:endpoint name="partURI" part="return"
  xpath="/p:Parts/Part/@xlink:href" interface="tns:partInterface"
  binding="dynamic"/>

  @binding attribute can have 2 values, "static" (default) and "dynamic". When
  @binding is "dynamic", the runtime *may*, but has not to, try to retrieve a
  binding definition from the newly created URI (perhaps with an extra
  path/request parameter to indicate that it's not the representation which is
  requested).
  If dynamic binding discovery is not attempted/fails, a binding statically
  referenced (as shown in [1]) will be used.
  If dynamic binding is used then a wsdl doc is returned. This returned wsdl
  must contain a binding for a given interface (probably identified the way
  shown in [1]), and may/should also contain a <service> element. One issue
  here is that the returned wsdl has to know about portTypes/interfaces
  referenced in the original wsdl doc, and a such, these portTypes must be
  grouped in a separate WSDL doc.

  Thanks
  Sergey Beryozkin

  [1]
  http://lists.w3.org/Archives/Public/www-ws-desc/2003Apr/att-0088/R085-2003-0
  4-22.html