[Pgi-wg] Sec: Agreement on WS-Naming (ref by strawman)

[Aleksandr Konstantinov]

On Friday 20 March 2009 12:21, Morris Riedel wrote:
> Hi Aleksandr,
> 
> >-IMO the real question is what such EPRs may be used for. If they planned
> as
> >-replacements for URLs and *full* such EPR is going to be put into
> >-every SOAP header then I wohuld vote strongly against. But if such EPR is
> just a way
> >-to describe service (same as corresponding section in Glue
> >-document) and are ony meant to be used in some kind of information
> system/port
> >-then why not.
> 
> Just for clarification: With URLs you refer to the wsa:To element of
> WS-Addressing correct? If so, they are of course in any SOAP message - or I
> didn't get your point.

With URL I mean URL which was supposed to identify resource.
Now with introduction of EPRs we get another layer of indirection in resource
identification. That's alright. But EPRs can also be used to describe resource.
And that description part may become lenghty. So if PGI goes for EPR I would
like to see it profiled which parts of EPR are expected/needed at which place/moment.
If my client application queries for some service with specific capabilities 
and gets 100kB EPR describing who where and how can contact service that is
probably ok. But if for contacting that service my application has to insert that
EPR into every SOAP message I would like to know which parts of it I can be safely 
removed in order to avoid buying dedicated optical link to computing center hosting
that service. :)  

> 
> Also, if it would be only to put into a kind of information system I would
> vote for crunching the EPR in elements that can be probably better searched
> for by an query instead of putting information into EPR.

I just have to agree.

> 
> Here, WS-Naming, provides basically a name of a service while its EPR might
> change over time (renewable reference). Also the document refers to
> trackability of messages once you use WS-Naming (e.g. UUID dependency) - so
> I expect that this information of WS-Naming EPR additions should be part of
> any SOAP message. But that's only my interpretation from the document.
> 
> 
> Q: Can we agree that each functional service in the EPR space such as
> OGSA-BES are contacted with an pure EPR using URIs? We don't consider
> accessing a BESService simply with an URI - or do we?  I would not vote for
> that.

I would like to know what do You mean by "contacted". If that implies human 
user is expected to write EPR by hand to make his/her client to contact some 
service then I disagree.


A.K.