Fwd: [MOBY-l] Which machine deals with ontology issues?

[mwilkinson at gene.pbi.nrc.ca]

Chris Wroe wrote:

> My current view is that the directory component (or mioBoby central) should
> perform the expansion of the query to include services with more general
> input types and not those with more specific types. If they are like me,
> neither service providers or users will think of doing this on their own
> and shouldn't have to.

This was pretty much the conclusion that we came to in our discussions as well.
We were also considering the "cost of implementation" of these ideas on the
server-side, given that one of the Ten Commandments of BioMOBY is that the effort
to create a service should be minimal to encourage participation.  We do, however,
want to make this an optional thing... e.g. include a flag in the parameters sent
to MOBY-Central to say "give me ONLY the services that deal with this EXACT object
type".  That way, it can be up to the intelligent client program to do an expanded
search only when it needs to.


>
> The second question is about a service accepting only one kind of data type
> or *any* kind of data structure that contains that type.
> myGrid took the simple option. Services only accept the simple data type
> and will fail when presented with that data embedded in some more complex
> data structure. The current pre-prototype workflows included some minimal
> format transformation steps to ensure the correct fragment of data was
> passed to the next service.

Here we differ just a little bit.

In day-to-day life, if people are actually using the registry and a
properly-written Client program, the server should never get objects that it can
not deal with.  It is only when someone hard-codes an end-run around the WSDL that
a server will get the wrong object types, or deprecated object types... and I
don't think we necessarily should be pandering to that audience ;-)

We concluded that the Service should not be *required* to validate objects that it
receives, and at most we should have an "ISA" call to MOBY-Central so the Service
can ask if "this thing" IS A "something I know" ... if it really wants to do so.
Generally speaking, however, it should simply parse the object on the assumption
that it has received something that is, or is a child of, the object type that it
understands.  Given the way we are implementing inheritance, this should succeed.
MOBY-Central and Client should ensure that it does, in fact, receive this type of
object.  If the input object is incorrectly formatted, the server IMO is justified
in puking out an error - tough luck!

It becomes somewhat more difficult when we move to "HAS A" relationships, since
you don't necessarily know whereabouts in the larger object to look for the
component that you understand...  in any case, we can deal with this later...


> The object ontology is planning for the future though! The hope would be
> that various components of the myGrid framework will in the future be able
> to interpret both ontological and XML schema datatype metadata to skim off
> appropriate elements and pass only those to the service.

yup!

For the moment, I am re-working the existing MOBY Central, and MOBY Client CGI so
that it operates according to the rules stated above.  It doesn't have to stay
this way - after all, this is a research project :-)   We're just tryin' stuff out
to see if it works.  If it doesn't, then we'll try it another way.

cheers all!

M

--
--------------------------------
"Speed is subsittute fo accurancy."
________________________________

Dr. Mark Wilkinson, RA Bioinformatics
National Research Council, Plant Biotechnology Institute
110 Gymnasium Place, Saskatoon, SK, Canada

phone : (306) 975 5279
pager : (306) 934 2322
mobile: markw_mobile at illuminae.com