ABI CellML Meeting Minutes, 16th June 2010

Present: Andrew Miller, Randall Britten, Justin Marsh, David (Andre) Nickerson, Tommy Yu, Poul Nielsen.
Apologies: Peter Hunter, Mike Cooling

  • Last week's action items:
    • Last week's action item 1 - Randall to send an email about the xml:base issue for Justin to investigate. This was done. The issue is fixed in the head revision, but not in OpenCell 0.7. One suggestion was to try again ignoring XML base. Consensus was that two things need to be done: it should be a curation guideline that models not include an XML base. OpenCell should warn users if a model doesn't work, and it has an XML base set. Action item 1: Justin to e-mail Catherine to ask her to update the curation guidelines. Justin also to add a tracker item to remind us to improve error messages in OpenCell.
    • Last week's action item 2 - Tommy to create a tracker item regarding a problem with FieldML exposures: CellML specific aspects should not be present on FieldML exposures, e.g. curation stars for various CellML software. Tommy has now created the tracker item, and has now fixed the issue as well.
  • CellML Core Specifications: Poul noted that he has been working on the CellML file with fundamental physical constants he created a couple of years ago. He noted that the first version used the initial_value attribute, and so doesn't imply that the value remains constant with respect to time. He has changed the file to use MathML equalities instead. He noted that the files are now very verbose. He suggested that adding a value attribute, alongside initial_value, might provide useful syntactic sugar. Andrew noted that flagging it as a constant which doesn't change from the initial value could be another option. Andre asked about constants in which there was uncertainty - consensus was that just because a parameter value specified is an estimate doesn't mean that it isn't a constant - or that a sensitivity analysis can't be done on the constant. Andrew noted that ModML will solve the verbosity issue without adding extra syntax, because it has a simple core, but Haskell can be used to succinctly express values which are transformed to a core expression. Justin noted that CellML is an intermediate format, and was never designed to avoid verbosity anyway. Consensus was that there is no need to change CellML as MathML is sufficient.
  • OpenCell:
    • Justin has been working on allowing IDA support from within OpenCell. The Saucerman model appears to be terminating at the wrong end time. Andrew suggested the possibility it was due to an out-of-date file generated from the interface.
    • Randall has organised a meeting about developing using Scala and the JVM, at 3 PM today (Wednesday)
    • Justin has received a number of reports of people on Mac OS X 10.5 having problems. OpenCell is known to work on 10.6, but Justin doesn't have access to a non-dev machine with 10.5 on it at present. Action item 1a: Justin to create tracker item for Mac OS X 10.5 issue
  • CellML API:
    • Andrew noted that there are currently problems with snapshot building on Windows, where the buildbot buildslave on Windows keeps disconnecting from the buildmaster. He has explored many avenues in an attempt to fix this - including upgrading buildbot masters and slaves, and hasn't found the underlying cause yet. It is possible that it is related to the VM technology changing something, but this has not been proven yet.
    • Andrew has changed CeLEDSExporter to allow for code which uses an implicit solver like IDA to be generated. Randall has experimented with this code, but has found that as the MATLAB mass matrix solver doesn't use a typical implicit DAE formulation, it doesn't look CeLEDSExporter can generate code which will work with this solver out of the box.
    • Andrew has been working on fixing an issue with the code generation (or CeLEDS Exporter) for implicit code which is causing a crash on Windows, with no problems on any other platforms.
    • Andrew fixed a bug in the code generation which was stopping initial value computations from working correctly, and has added a test for this case.
  • Repository contributions: No updates this week.
  • PMR2:
    • Tommy has made and staged release candidate 2.
    • It no longer puts curation stars for CellML software on FieldML models.
    • Randall asked why the FieldML view did not have as many navigation options as CellML models. Tommy said that this was because CellML Exposures take legacy 'tempdoc' data from models, but he wanted to move towards exposures being a listing of models, and data being in specific files. There was extensive discussion of this issue, and whether exposures should be richer or should simply be a cache / index of the Mercurial repositories/workspaces. Randall noted that the release demonstrates the use of non-CellML exposures in the repository, which is enough for this release, and so doesn't think any changes are required before the release.
    • There was a brief discussion of the papers about PMR2. There was consensus that it made sense to go ahead with the short application note and the paper on DVCS for building repositories rather than holding them up for a longer software article. Action Item 2: Tommy to send the application note around to all authors who haven't seen it yet.