[ogsa-wg] Re: [ogsa-naming-wg] WS-Names and WS-Addressing WSDL Binding
Frank Siebenlist
franks at mcs.anl.gov
Tue Oct 11 12:57:55 CDT 2005
I'd like to go back to the start of your discussion and talk about the
required properties of the Address.
The assumptions for the AbstractName are:
* The name MUST be globally unique in both space and time.
* The name conforms to URI syntax ("Uniform Resource Identifiers
(IRI): Generic Syntax", RFC 3987).
I believe that we need more assumptions on the binding properties
itself, and an example may help.
Suppose we have:
<wsa:EndpointReference
xmlns:wsa="http://www.w3.org/2005/03/addressing"
xmlns:name="http://ggf.org/name">
<wsa:Address>
http://tempuri.org/example?Id=3
</wsa:Address>
<name:AbstractName>
urn:guid:B94C4186-0923-4dbb-AD9C-39DFB8B54388
</name:AbstractName>
</wsa:EndpointReference>
As you can see, we have an AbstractName that clearly looks unique in
space and time, and an Address that seems to have a service/resource
identifier (Id=3) that doesn't look that strong.
Now suppose that this service application is implemented such that after
the hosting environment is recycled/restarted, the following EPR is valid:
<wsa:EndpointReference>
<wsa:Address>
http://tempuri.org/example?Id=4
</wsa:Address>
<name:AbstractName>
urn:guid:B94C4186-0923-4dbb-AD9C-39DFB8B54388
</name:AbstractName>
</wsa:EndpointReference>
(I've removed the name space identifiers for clarity)
As you can see, the implementation has assigned a different Address to
the same AbstractName, and we can guess that the Id=4 is the changed
distinguishing parameter value used for dispatching to the same
service/resource.
Now suppose that this same implementation will reuse the previous Id=3
for a different service/resource, like:
<wsa:EndpointReference
xmlns:wsa="http://www.w3.org/2005/03/addressing"
xmlns:name="http://ggf.org/name">
<wsa:Address>
http://tempuri.org/example?Id=3
</wsa:Address>
<name:AbstractName>
urn:guid:B94C4186-0923-4dbb-AD9C-39DFB8B77777
</name:AbstractName>
</wsa:EndpointReference>
So now the service implementation has reused the old Address with Id=3
for a new, different service/resource that is uniquely identified with a
new guid that ends with 77777 instead of 54388: we have two different
AbstractNames for two different services/resources that are bound to the
same Address, which could cause some confusion...
One could claim that this is a "bad" implementation and that the
EPR-minter should *ensure* that this should not happen. However, this
application could work perfectly correct for the things that it's
supposed to do, and it is the mismatch in assumptions that causes the
problem.
I believe that if we have a strong uniqueness claim for the
AbstractName, then the binding to an Address is only meaningful if that
Address (or the binding) has an equivalent strong uniqueness property.
In other words, the binding of the AbstractName that is "globally unique
in both space and time" to an Address, only makes sense if that Address
has the exact same property.
Now the only one that can ensure that the Address is also "globally
unique in both space and time", is the implementation, i.e. the
EPR-minter, and those implementations that could be identified through a
profile for EPR-minters: the
globally-unique-Address-in-both-space-and-time EPR profile.
So, binding a "globally unique in both space and time" AbstractName
would only make sense if one is assured that the implementation conforms
to the globally-unique-Address-in-both-space-and-time EPR profile.
One could relax this requirement a little by considering the complete
binding instead of the Address only, and we could for example consider a
validity time-interval and a binding issuer. This would allow you to
localize the uniqueness requirement of the Address to the time-interval
or/and issuer. But so far we have punted the idea of dealing with an EPR
as a true assertion, which leaves us only the Address to play with.
Summary: I believe that the "globally unique in both space and time"
requirement for the AbstractName MUST hold true also for the Address to
which it is bound, and that EPRs with such an Address constitute a
globally-unique-Address-in-both-space-and-time EPR-minter profile. One
can only bind AbstractNames to EPRs that adhere to the
globally-unique-Address-in-both-space-and-time EPR-minter profile.
Note that if this all holds, then the Address of an EPR that conforms to
the globally-unique-Address-in-both-space-and-time profile, could also
be used as an AbstractName, which would be great as it would give us an
identifier that we could use to bind (authorization-) policy to and as a
side-benefit would also be available on the wire.
Regards, Frank.
Maguire_Tom at emc.com wrote:
>
> As promised at the F2F in Boston I have started a thread of discussion on
> the subject line. I have reposted the thread to the this mailing list
> (ogsa-wg) in the hope that broader distribution will spur debate and
> discussion.
>
> Tom
>
> -----Original Message-----
> From: Maguire, Tom
> Sent: Sunday, October 09, 2005 4:51 PM
> To: Maguire, Tom; David.Snelling at UK.Fujitsu.com
> Cc: ogsa-naming-wg at ggf.org; tuecke at univa.com
> Subject: RE: [ogsa-naming-wg] WS-Names and WS-Addressing WSDL Binding
>
> Dave you mentioned in one of your question:
>
>
>>> It appears that in the example that either the was:Address and the
>>> soap:address must be the same or that the wsa:Addess is irrelevant.
>>> I can't really believe the former so let's assume the later.
>>>
>
> It's not that the wsa:adddress is irrelevant it is that the wsa:address is
> logical as opposed to physical. This is precisely why I think we can use
> it....
>
> Tom
>
> -----Original Message-----
> From: owner-ogsa-naming-wg at ggf.org [mailto:owner-ogsa-naming-wg at ggf.org] On
> Behalf Of Maguire, Tom
> Sent: Sunday, October 09, 2005 8:35 AM
> To: David.Snelling at UK.Fujitsu.com
> Cc: ogsa-naming-wg at ggf.org; tuecke at univa.com
> Subject: RE: [ogsa-naming-wg] WS-Names and WS-Addressing WSDL Binding
>
> Dave,
>
> I'll do my best to answer your questions inline below. Let me caution this
> thread a bit. The WSDL Binding specification is not complete and is clearly
> still evolving...
>
>
>>> It appears that in the example that either the was:Address and the
>>>
> soap:address must
>
>>> be the same or that the wsa:Addess is irrelevant. I can't really
>>> believe
>>>
> the
>
>>> former so let's assume the later.
>>>
>
> Yes, I believe that is the intent. As I mentioned in my note it is
> 'interesting' that they are the same. My guess is that makes
> implementations that are not <wsa:metadata> aware able to cope. I would
> expect that would be a 'best practice'. Not sure what the implications
> would be for us if that were the case...
>
>
>>> With a wsdl11:definitions section present, the wsa:Address field must
>>> be
>>>
> superseded
>
>>> by the soap:address chosen by the client. I assume that the
>>> soap:address gets copied to the was:To field in the soap header.
>>>
>
> Ultimately you are correct however I expect that the specification of that
> linkage would not be quite as explicit as that.
>
>
>>> There is no linkage in the wsdl11:definitions to connect the
>>> wsa:Address
>>>
> to it.
>
> No
>
>
>>> Q1) What happens with more than one wsdl11:definitions section in the
>>> was:Metadata?
>>>
>
> I have no idea what that would even mean. I presume they would limit that
> in the spec. As I said it is still evolving.
>
>
>>> Q2) In this case can we put any old junk in the wsa:Address?
>>> i.e. leave it out (except that the scheme saus [1..].
>>>
>
> <wsa:address> is required and I would assume that at a minimum there would
> be a statement of 'best practice' where the <wsa:address> is the 'default'
> address.
>
>
>>> Q3) If we use the wsa:Address as an Abstract Name, how do we know that
>>> is
>>>
> what
>
>>> we are doing? We could subtype the EPR to create a WS-Name as we do
>>> now, and bind the usage of the was:Address to type of the WS-Name.
>>>
>
> I would use a wsi conformance claim on both the wsdl and the EPR. The wsdl
> claim would be that the service is capable of generating WS-Names. The EPR
> claim would be that this EPR adheres to the additional semantics of a
> WS-Name.
>
>
>>> Q4) I thought WS-Addressing was NOT about naming or identity.
>>> How will this use (abuse) of the wsa:Address go down with the W3C folks?
>>>
>
> I think this is a misread on your part W3C objected to identity being
> encoded in something OTHER than a URI (IRI); in the WS-Addressing case they
> objected to ReferenceProperties. Ultimately ReferenceProperties were merged
> with ReferenceParameters which weakened (removed) the identity semantic. I
> think they would be extremely happy with the use of a URI as an identifier
> :-).
>
>
> Tom
>
>
> On 7 Oct 2005, at 12:41, Maguire_Tom at emc.com wrote:
>
>
>> This will be a fairly long note to discuss the current incarnation of
>> WS-Naming Abstract Names. An Abstract Name has the following
>> properties:
>>
>> * The name MUST be globally unique in both space and time.
>> * The name conforms to URI syntax ("Uniform Resource Identifiers
>> (IRI): Generic Syntax", RFC 3987).
>>
>> Let's leave aside the first point, for the time being, and focus on
>> the second point. The abstract name is an IRI which is an
>> internationalized URI. Currently this means that a WS-Name abstract
>> name would look like
>> this:
>>
>> <wsa:EndpointReference
>> xmlns:wsa="http://www.w3.org/2005/03/addressing"
>> xmlns:name="http://ggf.org/name">
>> <wsa:Address>http://tempuri.org/example</wsa:Address>
>>
>> <name:AbstractName>urn:guid:B94C4186-0923-4dbb-AD9C-39DFB8B54388</
>> name:Abstr
>> actName>
>> </wsa:EndpointReference>
>>
>> There are several built in assumptions in this particular rendering of
>> an
>> abstract name. First, there is an assumption that the <wsa:Address>
>> is the
>> [destination] MAP of the EPR. Second, the AbstractName does not need
>> to flow on the wire when 'dereferencing' this EPR.
>>
>> It may be ok for the AbstractName to not flow on the wire. I will
>> leave that discussion to others. Let's focus on the first
>> assumption...
>> If you assume that the <wsa:Address> is NOT necessarily a physical
>> address
>> (URL) then it is essentially the same as an AbstractName minus the
>> "MUST be globally unique in both space and time" property described
>> above.
>>
>> This is essentially how 'Web Services Addressing 1.0 - WSDL Binding'
>> defines
>> a <wsa:Address>. An example from that specfication:
>>
>> <wsa:EndpointReference
>> xmlns:wsa="http://www.w3.org/2005/03/addressing">
>> <wsa:Address>http://example.com/fabrikam/acct</wsa:Address>
>> <wsa:Metadata>
>> <wsdl11:definitions targetNamespace="http://example.com/fabrikam"
>> xmlns:fabrikam="http://example.com/fabrikam"
>> xmlns:abc="http://www.abccorp.com/"
>> xmlns:soap="http://schemas.xmlsoap.org/wsdl/soap/"
>> xmlns:iiop="http://www.iiop.org/"
>> xmlns:wsdl11="http://schemas.xmlsoap.org/wsdl/">
>> <wsdl11:import namespace="http://example.com/fabrikam"
>> location="http://example.com/fabrikam/fabrikam.wsdl"/>
>> <wsdl11:import namespace="http://www.abccorp.com/"
>> location="http://www.abccorp.com/abc.wsdl"/>
>> <wsdl11:service name="InventoryService">
>> <wsdl11:port name="ep1" binding="abc:soap-http-binding">
>> <soap:address location="http://example.com/fabrikam/acct"/>
>> </wsdl11:port>
>> <wsdl11:port name="ep2" binding="abc:iiop">
>> <iiop:address location="..."/>
>> </wsdl11:port>
>> </wsdl11:service>
>> </wsdl11:definitions>
>> </wsa:Metadata>
>> </wsd:EndpointReference>
>>
>> And also from 'Web Services Addressing 1.0 - WSDL Binding'
>>
>> In particular, embedding a WSDL service component description MAY be
>>
>
>
>> used by EPR issuers to indicate the presence of alternative addresses
>> and protocol bindings to access the referenced endpoint. The
>> alternatives are provided by the different endpoints of the embedded
>> service.
>>
>> It is interesting to note that in the above example that the
>> <wsa:address> matches the soap:address location.
>> So this says to me that the <wsa:address> is essentially equivalent
>> (or at least could be) to an abstract name.
>>
>> Thoughts?
>>
>>
>> Thanks,
>> Tom
>>
>> Senior Technologist, CTO Office
>> EMC²|SMARTS
>> 44 South Broadway
>> 7th Floor
>> White Plains, NY 10601
>> Office: +1-914-508-3477
>> Mobile: +1-845-729-4806
>> Email: maguire_tom at emc.com <mailto:maguire_tom at emc.com>
>>
>> If you want to build a ship, don't drum up people to collect wood and
>> don't assign them tasks and work, but rather teach them to long for
>> the endless immensity of the sea.
>>
>> Antoine de Saint-Exupery
>>
>>
>>
--
Frank Siebenlist franks at mcs.anl.gov
The Globus Alliance - Argonne National Laboratory
More information about the ogsa-wg
mailing list