[ogsa-wg] RE: Modeling State: Technical Questions

[Karl Czajkowski]

On Apr 05, Roger Menday loaded a tape reading:
> On Tue, 2005-04-05 at 17:09 +0700, Karl Czajkowski wrote:
...
> > As an example, I find it exceedingly strange that the job's transient
> > state is part of its structured name below, but its unique numeric
> > identifier is used as a reference parameter. To me, it should clearly
> > be the other way around. 
> 
> So, each job has it's own URI, eg, http://thegrid/users/ian/jobs/435
> 
> Then, doing a GET on this returns some document of job metadata
> 
> (??)
> 
> Roger
> 

I suppose, or really

  http://server.somewhere.grid/jobsXYZ/TlkUjg13DunN8ie2

I think the URI is really just an opaque dispatch path that makes
sense to the endpoint designer and not necessarily anyone else. As
soon as you start trying to organize or decompose this space according
to conceptual interface rather than implementation, you run headlong
into the community modeling tar-pit.

My starting assumption is that there will be different communities or
sects with different domain models. The danger with using URIs and
reference parameters for this is that it is very poorly typed, so
there is very little tooling support to express how the concept space
has been modeled nor to perform meaningful validation at
runtime. There is also very little support for building discovery
systems to catalog and organize information about different resources
using such an implied model. Using something like XPath or XQuery to
select resources from a virtual XML "group" document seems a lot more
disciplined to me, as does having XML representation of the "valid"
operations and metadata for those resources (WSDL and
WS-ResourceProperties, in the WSRF approach).

As to whether GET on resource URIs actually does anything useful, I'll
leave that for people who care about such things. :-)

Similarly, whether DNS is sufficient for practical, efficient,
resilient naming of URI namespace "authorities" is also something I'll
leave to others. Personally, I hope that the resource URI is more than
a transient service address.  It seems to me that some of the
WS-Addressing debate relates to whether this is the case in practice
or not.


karl

-- 
Karl Czajkowski
karlcz at univa.com