[MOBY-l] Re: [MOBY-dev] Semantic Moby and Web Services <-> Semantic Web integration efforts?

[Phillip Lord]

>>>>> "Catherine" == Catherine Letondal <letondal at pasteur.fr> writes:

  >> OWL-S/WSMO are trying to do a different thing to moby-s. The main
  >> aim for these technologies is to use highly expressive logics to
  >> enable automated composition and choreography of web
  >> services. 
  >>      
  >> Moby-s on the other hand takes a more constrained view of
  >> semantics. It's trying to produce semantic descriptions which are
  >> good enough to reduce the problem for the biologist to selecting
  >> from one or a short list, rather than then 500.
  >>

  Catherine> I meant S-Moby in fact, not Moby-S, sorry for this
  Catherine> mistake.

Ah, well, this makes quite a difference!

  >> There is a paper in this years ISWC (currently going on now!),
  >> which you can get from here...
  >>
  >> http://www.cs.man.ac.uk/~phillord/download/publications/biomoby-
  >> comparison-iswc2004.pdf
  >>

  Catherine> Thanks a lot for this information. I have read the paper,
  Catherine> which indeed helped a lot to understand your underlying
  Catherine> choices (BTW, I would be interested to know about the
  Catherine> feedback it had at the ISCW conference - generally,
  Catherine> bioinformatics is considered as a "very specific" area,
  Catherine> and since the differences between "generial" applications
  Catherine> such as e-commerce and e-science are very well explained
  Catherine> in your paper, I'm curious to know whether the audience
  Catherine> reacted).

It's difficult to say at the moment. I got some interest immediately
after the talk. The paper was written to be slightly
provocative, and, indeed, some people asked about my statement that
"automated service composition is (largely) inappropriate in
a scientific domain". 

Sadly, I seem to have picked up a illness at the conference, and was
in bed for most of the social events and poster sessions. This meant
that a) I didn't get a chance to talk to more people and get better
feedback and b) I missed some free beer, good food and Japanese
drumming. Those who of you who know me, will realise how irked I was
at this. 

So, I'll have to wait a while for people to read the paper, before I
am sure of the reaction.


  Catherine> I still have a question about the comparison to OWL-S. In
  Catherine> section 7 of your paper (Service provision and Service
  Catherine> interfaces), you explain that OWL-S defines a grounding
  Catherine> ontology, and that the 3 approaches (moby-s, s-moby and
  Catherine> myGrid) have chosen a simpler approach. 

The "grounding" in OWL-S is specifically that part of the ontology
which relates between the domain concepts describing the functionality
of the service (in the "process" and "profile" subontologies) and the
specific invocation interface, which for web services pretty much
means the WSDL document. 


  Catherine> I understand that the problem that is addressed here is
  Catherine> to have a high description not only of domain concepts,
  Catherine> but also on their relations, and that this is not
  Catherine> required in the 3 approaches, for different reasons. 

  Catherine> But I don't understand your explanations: How is the
  Catherine> S-Moby interface defined by its "upper" ontology? I have
  Catherine> carefully read the S-moby design document
  Catherine> (http://www.biomoby.org/S-MOBY/doc/Design/S-MOBY_Design_Overview.html),
  Catherine> what do you call an upper ontology there - the set of RDF
  Catherine> graphs submitted by providers? 

S-moby is basically a REST architecture, with RDF and OWL
messages. Strictly, it's service interfaces are all very simple, and
all identical; it's HTTP, post, get, and the other three commands that
I (and most people) can never remember. 

The response to these is, again, always the same for an s-moby
service, and defined in the spec. So you don't need an ontology to
describe this. There is no heterogeneity. 

Alternatively, you could consider the service interface to be specific
RDF graphs that an s-moby service returns. Perhaps this is the better
analogy, as this defines the specific functionality of the service. 
In this case, the graph defines the functionality totally. If you have
the graph, you know how to address the service, what queries to send
to it, and how to interpret the result coming back. So, again, you do
not need to ground a s-moby service to an underlying WSDL. k


  Catherine> You also mention that moby-s interfaces are not
  Catherine> heterogeneous (and thus does not need an ontology to
  Catherine> interconnect concepts). What is not heterogeneous and
  Catherine> why?

All moby-s services are the same. They have a single method, which you
can invoke. Again, if you have the ontological description, you can
understand what data you should send to a service, and how to
interpret the data that comes back. So, again, no grounding is
required. 

MyGrid is most like moby-s, in this context, except that instead of
having a single defined service interface, we have a couple, and a
pluggability layer within the other components to cope with the
difference. This is why taverna and freefluo can cope with, for
example, moby-s services. 


I hope this makes sense. It's quite hard to describe, particularly as
s-moby's REST interface breaks (my use of) the standard web services
terminology somewhat. Let me know if things are unclear. 

Cheers

Phil