Resolving endpoints

[Paul Harrison]

I fully endorse Guy's original point on this, that the current accessURL 
situation is ambiguous when it comes to web services and WSDL. I think 
that as a minimum we should have a separate regisitry element that 
points to the WSDL (or add wsdl as an option for the use attribute ).  
So we might have the following snippet of WSDL service definition

    <service name="test">
        <port binding="tns:MyServiceBinding" name="MyService">
            <soap:address location="http://localhost:8000/ccx/binding1"/>
        </port>
        <port binding="tns:HeartbeatBinding" name="HeartBeat">
            <soap:address location="http://localhost:8000/ccx/test"/>
        </port>
    </service>


there would be a registy entry like
        <Interface>
            <Invocation>WebService</Invocation>
            <AccessURL 
use="wsdl">http:www.astrogrid.org/astrogrid-applicationsCommonExecutionConnectorService.wsdl</AccessURL>
            <ServicePort>MyService</ServicePort>
        </Interface>

Where the entry for <ServicePort> points to the name attribute of the 
<port> in the wsdl above. If we were to insist that the standard 
services had standard names, the the <ServicePort> element could just 
point to the  "main" service  that  was being offered - otherwise we 
will have to come up with a way of saying in the registry which  each of 
these are - possibly with multiple <servicePort> entries with attributes.

        <Interface>
            <Invocation>WebService</Invocation>
            <AccessURL 
use="wsdl">http:www.astrogrid.org/astrogrid-applicationsCommonExecutionConnectorService.wsdl</AccessURL>
            <ServicePort type="main">MyService</ServicePort>
            <ServicePort type="heartbeat">HeartBeat</ServicePort>
        </Interface>

This means that it is a more complicated to look up the endpoint for a 
particular service, but this complexity is forced upon us by the 
flexibility that is allowed by web services and WSDL. As Guy also 
suggested it might be a good idea to have the lookupEndpoint() function 
as part of the standard registry interface, so that most clients do not 
have to worry about this complexity.

Paul.

Ray Plante wrote:

>Hi Guy,
>
>There is still ambiguity about what AccessURL points to in the case of Web 
>Services.  In VOResource v0.9, it is defined to point to the WSDL; 
>however, there was some desire for it to point to the endpoint for the 
>reasons you described.  In the NVO registry, we didn't have a lot of 
>Web Services defined (and that we critically depended on for our 
>apps/demos), so it we were living with this ambiguity while we figured out 
>what made sense.  
>
>There are still some issues to address related to our discussion of 
>standard WSDL definitions that will affect this question.  For example, 
>would we need or allow different parts of the interface to have different 
>endpoints?  If so, which one should be the "accessURL"?  Should the other 
>endpoints be placed in the registry as well? (probably so; then 
>additional metadata are needed.)   I'm doing some experiments now with 
>Axis to test some of the ideas we've batted around.
>
>cheers,
>Ray
>
>
>  
>

-- 
Dr. Paul Harrison, Astrogrid Developer
MERLIN/VLBI National Facility, University of Manchester,
Jodrell Bank Observatory, Macclesfield, Cheshire SK11 9DL, U.K.
tel +44 (0)1477 572681 (direct), 571321 (switch) - 07904025192 (mobile).