[MOBY-l] Objects used for discovery

[Ken Steube]

Hi all...there were two or three discussions during the meeting that I
have thought about, and I'm beginning to think we're going down a bad
road.  Let's try to flesh them out here.

First topic:

We had a brief discussion about input objects and their use for
discovery.

The issue was:  given a sequence I want to be able to discover all I
can do with that sequence including BLASTing it, feature analysis, and
anything else anyone ever thinks of.  But then in order to simplify
the handling of inputs and outputs I suggested having a BLAST input
object which extends a sequence object to include the usual BLAST
parameters.  Feed it into any BLAST service and you'll have everything
needed to run BLAST (including your sequence).

But now this issue arose:  If you try to discover with a BLAST object
will you get the same thing you'd get if you tried to discover with
just a sequence object?  Since you've specialized your sequence object
to include BLAST parameters you might miss some of the other services
that might be discovered.  But in my opinion by going from a sequence
to BLAST input you should have broadened the types of services you can
discover.  You still have the sequence and can discover feature
searches and anything else, but now you have other bits as well and
can whatever else is allowed with those bits such as BLASTing.

I think we can really make MOBY discovery more powerful than the
current design.  All you have to do is to discover based on the BLAST
input object and all the objects that share an ISA or HASA
relationship with it. This is fairly simple do implement as long as
our discovery is a logical OR operation for input object types.

Second topic:  later!

Ken

-------------------------------------
Ken Steube            steube at sdsc.edu
San Diego Supercomputer Center @ UCSD
San Diego, California             USA