[ogsa-hpcp-wg] MS HPCBasicProfile Web Service

[Peter G. Lane]

Chris tells me that this was discussed at one of the calls I missed, so I took a look at what the 
issues are for us. I can probably implement a singleton resource that acts as a default factory 
resource so that clients don't have to implement the full WS-Addressing spec. I've been told that 
our service framework should interpret this as a stateless service call and be ok. And since 
everybody else is using URLs, you'll probably just ignore any WS-A markup that might come in with 
the SOAP message created by our clients.

The caveat is that no operations will be available on the Activity. I think this was assumed 
already, but I just wanted to make sure for certain. All Activity manipulation will be through the 
factory. Is this correct?

One last concern was the version of the WS-Security Username Token spec. Was there agreement on 
which version everyone should use?

Thanks, and I think we're finally ready to commit to doing this. Hopefully I can get something 
working in a week or so.

Peter

Peter G. Lane wrote:
> Christopher Smith wrote:
>> Can I be clear on something?
>>
>> Instead of:
>>
>>     https://aristotle.dreadnought.org:9090
>>
>> You want:
>>
>>     <wsa:EndpointReference 
>> xmlns:wsa="http://www.w3.org/2005/08/addressing">
>>         <wsa:Address>https://aristotle.dreadnought.org:9090</wsa:Address>
>>     </wsa:EndpointReference>
>>
>> ... for my endpoint, at any rate. I'm guessing that yours will require 
>> me to
>> set some SOAP header blocks based on the WS-Addressing SOAP binding. I 
>> had
>> hoped to avoid this detail for now....
> 
> I'm not sure what else I can tell you. We require EPRs for interop and I 
> expected this not to be a problem since it's being used in the spec. I 
> didn't think it was optional to support just enough of WS-Addressing to 
> be able to return an Activity EPR. Saying BES uses WS-Addressing to me 
> means we completely support WS-Addressing.
> 
>> By the way ... it's not anonymous https ... we are authenticating the 
>> server
>> with SSL.
> 
> Right, "https" means http over SSL. "Anonymous" means anonymous SSL 
> authentication. What am I missing?
> 
> Peter
> 
>>
>> -- Chris
>>
>>
>> On 14/10/06 10:47, "Peter G. Lane" <lane at mcs.anl.gov> wrote:
>>
>>> Michel Drescher wrote:
>>>> All,
>>>>
>>>> Peter G. Lane wrote:
>>>>> Let me elaborate on why I need EPRs to be a requirement. Our
>>>>> implementation will be a stateful service with a static state resource
>>>>> for each resource manager that we interact with on the back end (i.e.
>>>>> fork/exec, PBS, LSF, Condor, ...). What this means is that you have to
>>>>> have a resource reference that cannot be obtained from the WSDL in 
>>>>> order
>>>>> for our service dispatcher to know what state resource is associated
>>>>> with a particular CreateActivity (or any other BES-Factory operation)
>>>>> call. This is what EPRs are primarily good for. I realize stateless
>>>>> services don't need it, but if EPRs aren't supported then there's no
>>>>> realistic way we can participate seeing as thought we'd have to 
>>>>> rewrite
>>>>> GRAM to, I guess, only do fork/exec. I know .Net has support for EPRs
>>>>> (heck, Microsoft was a coauthor for the WS-Addressing spec!), so I 
>>>>> don't
>>>>> see this as a huge burden. And as I said before, we are assuming EPRs
>>>>> for the Activity references, so it's a bit puzzling that anybody would
>>>>> use anything else for the factory.
>>>> I strongly support Peter here on requiring EPRs.
>>>>
>>>> It really shouldn't be that difficult to define a simple (I mean
>>>> *really* simple!) factory WSDL having one method with no input that
>>>> returns exactly one EPR. Agreeing on a transport binding shouldn't be
>>>> that difficult either, shouldn't it?
>>> Not to look a gift horse in the mouth, but why are you suggesting a 
>>> method to
>>> retrieve the EPR? I'd
>>> still need an EPR to call an operation that returns the service's 
>>> EPR, so it
>>> doesn't make much sense
>>> to me. If there's no discovery mechanism, then I'd suggest just 
>>> writing the
>>> EPR to a log file and
>>> distributing it from that.
>>>
>>> As for a transport binding, I thought we already agreed on anonymous 
>>> https.
>>>
>>> Peter
>>>
>>>> Implementing that factory should be a finger excercise.
>>>>
>>>> Problem solved?
>>>>
>>>> Cheers,
>>>> Michel
>>>>
>>> -- 
>>>   ogsa-hpcp-wg mailing list
>>>   ogsa-hpcp-wg at ogf.org
>>>   http://www.ogf.org/mailman/listinfo/ogsa-hpcp-wg
>>
> 
> 
> ------------------------------------------------------------------------
> 
> --
>   ogsa-hpcp-wg mailing list
>   ogsa-hpcp-wg at ogf.org
>   http://www.ogf.org/mailman/listinfo/ogsa-hpcp-wg

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 3804 bytes
Desc: S/MIME Cryptographic Signature
Url : http://www.ogf.org/pipermail/ogsa-hpcp-wg/attachments/20061016/b5ee18d4/attachment.bin