Meeting Minutes 8 August 2007

Present: Poul Nielsen, Randall Britten, Alan Garny, James Lawson, Catherine Lloyd, Peter Hunter, Andrew Miller, Tommy Yu (speakerphone). Apologies: Matt Halstead

Alan is keen for there to be a user meeting where users of PCEnv and COR can give some indication about how they would like the PCEnv and COR philosophies to be merged and what features from COR are higher priorities for them. It was arranged for 2 pm, and Peter, Poul, James, and Jonna (together with Andrew and Alan) will be able to attend. Alan will produce an agenda / list of items to guide the meeting before then. At some point after this Alan and Andrew and others who want to attend will have a technical meeting to discuss what needs to be done at a technical level to implement this.

There was a brief discussion about the session loading feature in PCEnv. Peter was concerned about the warning message which comes up when you load a session from the website. Andrew says that this is because the command line mechanism is mainly designed for opening models, and the warning occurs when you try to open a session as a model file. Opening sessions is a potentially dangerous task, because sessions include a lot of data about the list of open models, the layout, graph data and so on which could potentially override data in the current session that the user has taken hours to create, so we need a warning if people are going to configure their browser so that arbitrary websites can open sessions in PCEnv. Peter agrees that there are some circumstances when the warning is appropriate, but in the most common case, people are opening PCEnv for the first time. It was agreed that we would set some rule under which the warning is suppressed because no data is lost (model name passed on the command line + no other models already loaded).

Randall noted that there has been a proposal to add 'flags' to models to describe whether they e.g. comply with the CellML spec / a best current practice document, whether they have been validated by a third party to be a correct representation of the model, and so on. Also, James is keen for there to be a link to a wiki page about model curation which provides further information that cannot be represented as flags. There was consensus that icons were better than ticks from a usability perspective. It was agreed that James should come up with a concrete proposal / discussion document and send it to the list, and that Matt, Andrew, Tommy, and anyone else who wants to be involved could comment on it.

Randall summarised an e-mail from Matt: * Matt is following up with Michael about the progress on sitetools. * Matt is working on writing up the outcomes of the repository redesign wiki in a document.

Randall is now favouring an approach of dealing with the tracker evaluation in one step, with everyone justifying their tracker evaluations through an indication of how they ranked the trackers and the criteria, which he will later be able to reconcile to get a decision.

There was a discussion about what to do next to get a CellML 1.1.1 specification. There are two things proposed, firstly the deprecation of reactions and secondly repairing various errata type issues such as typographic errors and other mistakes where the intent of the authors is expressed unclearly or incorrectly. There are several issues found by Andrew and Alan, and according to Alan, there are also some issues found by Jonathan Cooper, although we don't seem to have this in the tracker and Alan wasn't sure of the details.

There was a brief discussion about the models currently in the repository that cannot easily be fixed to not use reactions. James suggested deleting them, but the consensus was that we shouldn't do this. Models include some text-book models which didn't provide maths, and some complex models like Bhalla & Iyengar.

Poul thinks that it is important that we have a justification for the removal of reactions. Andrew raised the issue of where this non-normative material is kept. Poul suggested that we could put it up somewhere and link to whenever circulating the new proposal. Another option would be to put it in the spec - there is currently a lot of non-normative material such as examples and explanations in the CellML specification. Andrew and Poul agree that the non-normative material makes it hard to read the spec when you are looking for a normative definition of CellML, but that the examples and explanatory material are important. Poul suggested that we could provide a normative-only CellML specification, and then provide an annotated CellML specification (Andrew suggested we call it a tutorial) which includes the examples as well as explanatory material. This should include a justification of why reactions are moved, as well as a description of the alternatives.

Catherine suggested also adding the vector support and discrete delay support into 1.1.1, but the consensus was that this should wait for 1.2. James asked if we should announce the addition of these elements ahead of time, but Poul thinks there is a danger that people will start using 'pre-announced' features. Poul noted that a similar argument could be extended to reactions. Andrew noted his opinion that because CellML 1.x versions are inherently incompatible due to differing namespaces, there is no point having a deprecation process. Peter drew a comparison to FORTRAN specifications. Andrew noted that this case is somewhat different because most (but not all) valid programs in one FORTRAN specification are valid programs in a later specification too, while all CellML 1.0 models are invalid CellML 1.1 models. An possible alternative suggestion from Peter was that we draw attention to the change through the mailing lists or similar. The group felt that we could not make any decision without Andre's input. However, the possibility that we would not reach consensus as we are going over the same issues was raised. To resolve the issue, Poul agreed to make a final decision this time next week, to give time for Poul to hear opinions about both approaches. Andrew noted that he is not opposed to deprecating reactions, but he thinks that we need to justify it in the context of the process for advancing CellML. Peter suggested that Poul also decide on how to make this process consistent next week.