[MOBY-l] Biomoby and Taverna 1.2

[Tom Oinn]

markw at illuminae.com wrote:
>>If noone (except of Heiko Dirk and me) want the collections to be
>>changed in the BioMoby API we will have to change the default behaviour
>>of Taverna back to the old one!
> 
> 
> I think your use-case for wanting the behaviour of collections to change is
> unfounded, since it overloads the intended meaning of the Collection element by
> putting WAAAAAAY too much semantics in it.  The use case of having sets of
> multiple alignments represented as Collections of Collections means that the
> inner-most Collection now not only represents a "bag" of things, but a "bag" of
> things that **DO** have a close semantic relationship to each other - these are
> a set of related sequences, and a set of related sequences **is a useful
> Object** in the context of MOBY, could easily trigger the discovery of a
> particular set of services that can specifically operate on sets of closely
> related sequences (versus sets of unrelated sequences), and therefore should be
> given a class of its own (IMO).

True, but you probably also want to allow, for example, a GC percentage 
service to run over that set of sets and produce a set of sets of scores 
- if the set of sequences becomes some opaque type without this 
information then you can't do this. Strikes me that you need a more 
sophisticated typing scheme to do this, perhaps extending the Collection 
concept to allow the imposition of semantics to the group of child 
objects? I think you do need nested collections though, it's not 
impossible to deal with (we have some slightly twisted code in Taverna 
to do it but it works well enough)

Tom