RE: [wsdm] [UPlat] definitions

["Sedukhin, Igor S" <Igor.Sedukhin@ca.com>]

David,

1.

GSH's URI and a URI of a WSDL element are different by definition. What I meant was the URI of a <service name=".."> or <port name=".."> element in WSDL 1.1 sense. Those URIs would identify elements that must be unique by the semantic definition of the elements in the WSDL spec (normative uniqueness).

<descriptions name="abc" targetNamespace="urn:oo">
<service name="a">
<port name="b" binding=".."><soap:address location="c"/></port>
</service>
</descriptions>

So the URI "urn:oo#a/b" would uniquely identify a service endpoint with location c. There could be another endpoint with location c, but that would be a different endpoint by WSDL definition.

In general adopting a URI to identify a resource, as defined by W3C, is a good thing, IMO.

2. 

As in my basic definition the registration, discovery and location are merely a set of operations and schema associated with each activity. So they are all separate set of operations. Any component could implement location but not registration operations, for example. So the basic definition does not preclude OGSI approach in any way. When referring to Web services platforms, though, I used an example of a Registry, implying UDDI registry (of course). That appears to be more ubiquitous and understandable today, so that the definition can be adequately interpreted by a larger community of users and platform vendors. If I were to refer to Grid platforms, perhaps, I would have used OGSI interpretation of the basic definition.

-- Igor Sedukhin .. (igor.sedukhin@ca.com)
-- (631) 342-4325 .. 1 CA Plaza, Islandia, NY 11788

-----Original Message-----
From: David Snelling [mailto:d.snelling@fle.fujitsu.com] 
Sent: Friday, October 17, 2003 8:36 AM
To: Sedukhin, Igor S; wsdm@lists.oasis-open.org
Subject: Re: [wsdm] [UPlat] definitions

Folks,

More thoughts from the OGSI world.

On 16/10/03 22:41, "Sedukhin, Igor S" <Igor.Sedukhin@ca.com> wrote:

> Identification/Identity
> 
> Identification is a way to represent that one element is same or 
> different than the other without necessarily looking at the contents 
> or definition of the element. Applied to Web resources in general, as 
> defined by W3C, a Uniform Resource Identifier (URI) can be used to 
> represent the identity of a Web resource. Applied to Web services, 
> every element which Web service description is composed of (service, 
> interface, endpoint, location, etc.), is required to be identifiable 
> by a URI. Such URI must be unique by definition. Identity is not 
> required to be an addressable location, so a dereferencing mechanism may be required to actually locate the corresponding resource.
> 
> Manageable resources need to be uniquely identified for the manager to 
> tell one from the other and also to consistently refer to. Therefore a 
> standard representation of a unique identity is required.

In OGSI a Grid Service Handle (GSH) is a URI and for all time refers to exactly one service instance. However, if I have two different GSH's they MAY refer to the same service instance or different instances. We spent a lot of time coming to this position with several use cases for not requiring an isomorphic mapping.

OGSI does not mandate a particular handle scheme, only defines a set of properties, like above. Any handle scheme you devise would need to add isomorphic properties. The challenge then becomes defining "sameness". With physical resources, it is relatively easy, but with services things become more fun.

> Registration/Discovery/Location
> 
> Registration is a method of advertising an existence of an element so 
> that it can be discovered. Discovery is a method of locating an 
> existing element so that it can be used or operated. Discovery can be 
> based on a selection criteria or simply a name or identity of an 
> element. Location is a method of obtaining an address of an element. 
> For example, location may mean translating an identity of an element 
> into an address of an existing useable element. In the Web services 
> sense, registration, discovery and location can be represented by a 
> set of operations and schema which may be implemented by a Registry. A 
> Web service can register itself or can be registered by a third party 
> by sending a request to the Registry. A Web service can be discovered 
> by sending a request to the Registry. The Registry can return the 
> description of a Web service with location address included in a description or it may return the location address directly.
> 
> Manageable resources have to be discoverable by the managers. 
> Manageable resources exposed via Web services can be registered, 
> discovered and located via a Registry.

OGSI has these concepts split. Discovery from a Registry (not explicitly defined in OGSI) would provide a GSH. Location (a GSR) is then provided by a HandleReslover. However, the two can be combined operationally without causing any problems. I would separate the definition of Location out from the others though. 

-- 

Take care:

    Dr. David Snelling <d.snelling@fle.fujitsu.com>
    Fujitsu Laboratories of Europe
    Hayes Park Central
    Hayes End Road
    Hayes, Middlesex  UB4 8FE

    +44-208-606-4649 (Office)
    +44-208-606-4539 (Fax)
    +44-7768-807526  (Mobile)