ABI Meeting Minutes 2008-04-23

ABI Meeting Minutes 2008-04-23

Present: Randall Britten, Peter Hunter, Justin Marsh, Andrew Miller, Tommy Yu

Apologies: James Lawson, Catherine Lloyd, Poul Nielson

Review of last week's action items:

Action Item 1: Andrew contacted the Kyoto group about their DAE system isolation algorithm, and emailed the cellml discussion group. Alan has responded.

Action Item 2: Tommy has found a tracker item (TI #371) which was already being used for the discussion of parallel/embedded workspaces, and has renamed it to be more accurate.

Agenda for this week:

What to do about an RDF service?

  • Randall questioned whether the any rdf parser should be exposed, especially if it is intended as an internal implementation detail. If the RDF service were not exposed at all, then there would be no issue of changing or contaminating the interfaces of the api.

  • Andrew replied that an rdf parser would be useful in a number of different services in the CellML API, and for PCEnv.
  • With regards to creating contamination of the interfaces of the core api, it is envisioned that the rdf parser would usually be used to 'pull' from within a service, rather than 'push' across to a different service. If, however, rdf objects did need to be passed around between services, then some interface would likely need to be defined anyway.
  • Passing around strings of rdf would not allow for`liveness' of data, and would mean that boxing and unboxing would have to happen at each point of transference. Randall brought up the case of application builders who were already using an rdf library; would it make sense to make them have a dependancy on another rdf library as well? The use of several different rdf libraries would also not allow liveness of rdf between parts of the program using different libraries, and boxing and unboxing the rdf would have to be done at each point of transference.
  • Professor Hunter noted that there were several experiments going on with regards to rdf and metadata; Andre's work is one, DAE tagging is another. In the future, we would want a much larger usage of metadata.
  • Tommy enquired about possible support for SPARQL; Andrew noted that we do not intend to implement that at present. It may end up being implemented, or we may abstract what SPARQL would be used for through a higher level metadata service.
  • Action Item 1: Justin will send an email to cellml-discussion about the discussion around an rdf parser, and include a link to the tracker item (TI #358). The discussion will close at 5pm on the 7th of May, 2008.
  • It was suggested that, if we did proceed, we should provide flags to switch off the rdf service, as we have with the other services. This would also turn off parts of other services that had dependencies on the rdf service.
  • Randall also noted that we may benefit from formalising which parts of the api are community assets, and where in the code it starts being mere implementation details. It was suggested that the idl for all core services plus any idl level dependencies was `the api', and everything else was implementation.

PMR2 Progress update

  • Tommy has improved the indexing of exposed models. PMR2 now uses rdf more comprehensively, and Tommy has switched from 4suite to rdflib for dealing with rdf. The database back end is XML under version control via Mercurial
  • Tommy did some work to fix some buildout issues that had arisen.

  • Once the prototype is done, he can enter test driven development and get feedback from James Lawson, Catherine Lloyd and others using the prototype.

Other items

  • The internal architecture of PMR2 was discussed.
  • Professor Hunter brought up the usage of CellML by Bioinformatics groups; that it would most likely be useful as generated bridges across pathways with known reaction kinetics, in order to build a model out of a query to a pathway database.
  • Dan Beard's group in NCW is working on something like this, using Matlab as the generated model language; Professor Hunter said that their database schema is thus far relatively flexible, and we should look into it.
  • Matt, Sarala and Poul are also working on a similar idea.
  • Andrew intends to attend the `Birds of a Feather' parts registry workshop at the Synthetic Biology Workshop, to talk about the CellML Model Repository.
  • Action Item 2: Justin to make Windows snapshot of PCEnv, so that we can see how 'fixed' MathML appears.