Meeting Minutes 29 August 2007

Present: Jonna Terkildsen, Catherine Lloyd, Randall Britten, Alan Garny, James Lawson, Andrew Miller, Tommy Yu (speakerphone).

  • Jonna noted that the CellML 1.0 and 1.1 specifications say that there can only be one connection between pairs of components. PCEnv doesn't enforce this but COR does. Andrew thinks that we can't put this into CellML 1.1.1 because it would mean that there would be valid CellML 1.1.1 models which are not valid CellML 1.1 models. No one at the meeting objected to the idea of it being in 1.2, so the decision was to propose this formally for community consensus.

    • Action item: Andrew to propose inclusion in CellML 1.2 by putting it up on the bug tracker and notifying the community.

    • Action item: Andrew to file a tracker item so that when PCEnv gets validation support, it can also offer to correct CellML 1.0 / 1.1 models broken in this way automatically.

  • Alan noted that Andre (David Nickerson) wanted to import the same model twice but only have one instance in the sense that both imports would refer to components which maintain their connections. Andrew doesn't understand how this could work given that the most common use of imports is to create two different instances of something sharing the same mathematics, i.e. the exact opposite of what Alan said. It is possible that we were not fully describing the issue.

  • Alan has forwarded an e-mail from Jonathan Cooper containing errata for the CellML specification to the mailing list. The group decided to deal with this by e-mail first due to the complexity of working out exactly what the effects of the changes were in a meeting environment and due to uncertainty about the relationship to items in the current errata.

    • Action item: Alan will try to compare the errata proposed by Jonathan Cooper to the current errata and send an e-mail to the list.

  • There was a brief discussion about what we should do to put up action items. Randall thinks that we should not have too much red tape controlling how minutes are written up, and is happy with either inline action items as long as they stand out, or a separate section for action items.

    • Action item: Andrew will start adding action clearly demarcated action items to the minutes.

  • There was a brief discussion about how to deal with singularities such as division by zero in models. Alan noted that Jonathan Cooper's implementation dealt with the problem by using an epsilon value around divisions by zero and so on so e.g. A=1/(B-5) is equal to A=(1/(B+epsilon-5) + 1/(B-epsilon-5)) / 2. Andrew doesn't like the idea of automatically replacing normal mathematical behaviour at certain points with different behaviour, and thinks that model authors should define this themselves in the CellML. There was some concern that these cases may be unimportant and may clutter the model (hence why they are usually not published), but the consensus was that this isn't really a big enough issue to justify not describing the entire model in CellML.

    • Action item: Andrew will file a tracker item requesting PCEnv be given a debug mode of operation which will stop the integrator on singularities. This might end up being Linux specific, although there may also be a way to handle this on Win32.

    • Action item: Catherine will commence work on a best practice document for model authors which will warn them of this issue.

  • James and Catherine are currently not using versions and variants in exactly the same way. The plans for the new repository aim to contemplate this, so the current system is a compromise anyway. James and Catherine have already resolved this, so there is no need for further action at this stage.

  • James is producing an ICSB poster which Edmund Crampin has offered to take to the ICSB conference. He has to have an abstract in by September the 1st, which makes it a tight deadline.

    • Action item: James will write up the abstract and send it to the team mailing list for feedback.