ABI CellML Meeting minutes 2009-06-10

Present: Dougal Cowan, Andrew Miller, Mike Cooling, David "Andre" Nickerson, James Lawson, Julian Eden, Tommy Yu, Randall Britten, Abhishek Tiwari, Peter Hunter, Matt Halstead, Poul Nielsen

Apologies: Catherine Lloyd, Justin Marsh

Last week's action items:

1) Catherine will contact the Newcastle group and will request input from Sarala and Matt.
  • Catherine has contacted the Newcastle group.
2) James to promote IUPS workshop - email attendees, organizers, mailing list, and create a workshop web page.
  • James has sent an email.
  • Randall suggested revisiting the content of the workshop, suggesting a presented introduction to CMGUI and CellML, and then having the attendees choose which of these they wish to have a practical tutorial on.
  • Peter said that splitting the attendees into CellML and CMGUI groups sounds fine.
  • Randall asked whether we should recruit more of the ABI IUPS attendees to help at the workshop. Peter said yes, everyone from ABI at IUPS should be there.
Action item 1) James to get the IUPS website information on the workshop updated.

This week's agenda:

1) Bangalor workshop, 19-21 November - update
  • Randall said that Andre is probably going to be the ABI member going to the conference.

2) IUPS preparation
  • James thought that we should organize a meeting to sort out the tutorial.
  • Peter emphasized the need for good handouts.
  • James mentioned that he is making a resource web page for the tutorial - links, tools, information.

3) Units validation - which is right?
  • James asked which is correct on units validation, COR or OpenCell (ValidateCellML)?
  • James said that he has tried using JSim for unit validation, but that too many of the models do not load in JSim.
  • Andre suggested that ValidateCellML might be useful as a fully standalone product.
  • James reported discrepancies in the errors/warnings given by different versions of ValidateCellML, that is the API version vs. OpenCell.
  • Andrew said that a nightly build of OpenCell would provide the latest ValidateCellML.
  • Andre said that a stable build is required for curation.
  • Randall suggested that the repository could be annotated with the errors, and with a note saying that the errors reported by the software are incorrect.
  • Andrew said that the difference between OpenCell and COR should be resolved. He suggested that we discuss what the validator should do in these cases under scrutiny by James.
  • There was a discussion about what should be done in various cases where units were multiplied, such as;
Should OpenCell give a warning for these equations?Should OpenCell give a warning for these equations?
Should OpenCell give a warning for these equations?
  • Mike said that people often do this by accident, and a warning could be useful.
  • Matt suggested that this sort of thing will occur frequently with imports.
  • There was a discussion about what is valid or correct when multiplying units.
  • It was generally agreed that COR is correct, and that OpenCell should be changed.

4) PMR2 update
  • Tommy has PMR2 staged, and hosting is now available.
  • Tommy is creating mappings for migration, and will try to finish migration to workspaces by the end of the day.
  • Randall suggested that upon starting the sanity check for migration, that PMR1 be frozen, assuming deployment of PMR2 within 2 days of the check starting.

5) Website redesign - update
  • James reported that at the meeting yesterday, it was decided to deploy with the new content instead of having the old content in place at launch.
  • Poul said that there were some questions on hosting reliability with internal hosting.
  • Peter said that he has seen no evidence of low uptime - he requested the logs of this information.
Action item 2) James to get the logs of cellml.org uptime for Peter.
  • Randall asked if it is possible to run both cellml.org and PMR2 on the same ITS server?
  • Tommy said yes, although we may need more storage at some point.
  • Randall said that we should just go ahead and host both separate Plone instances on the same server.
  • Tommy reported that Plone content version tracking is fixed in the latest version.

6) Repository contributions - update
  • James is still curating and updating statuses, updating e-notations one by one.

7) CellML Specs - update
  • At last week's CellML specifications meeting, it was agreed that we would not assign Andrew or anyone else to working on the functional-programming derived language for models, as that would take away time from more short-term activities. This could be a PhD project for someone, and might be sufficiently different to be called something other than CellML. Instead, we will work on ad hoc incremental changes for the time being.
  • Peter asked if any of Andrew's ideas could be used for FieldML. Randall said that the functional programming derived approach could form a new language which achieves the goals of both FieldML and CellML.
  • James said that metadata has largely stalled, due to time constraints as well as Tommy wanting a better representation of the metadata for PMR2. This might have to wait until after PMR2 0.1.
  • Randall called a breakaway on the core spec.

8) CellML API - update
  • Andrew has been working on setting up a new build system for the CellML API and OpenCell, using buildbot. He has created a buildmaster configuration, which Andrew Cantell has installed at http://autotest.bioeng.auckland.ac.nz:8010/ . There was some discussion about where the buildslaves would be. Andrew said that they are currently running on his development machine, but the plan is, once the buildslave VMs are set up, to run them on there. Randall noted that he had a discussion with Andrew Cantell about the Windows buildslave boxes, and due to high Windows licensing costs, the plan is now to use old desktop systems (which already have Windows licenses) rather than using a VM as for Linux. There was some discussion over whether we could cross-compile for Win32 to avoid these issues, but that raises questions about how different our test environment is to the end user's environment".
  • The buildbot buildmaster is configured for API, OpenCell, and Java builds, although not all build types have build bots set up to run them yet.
  • Randall suggested that we might want to look at doxygen; this is set up for CMISS and could be done for CellML.
  • Randall is trying to sort out a Windows build slave with Andrew Cantell.
  • Sarala has started writing documentation for the Java wrappers of the API.

9) OpenCell depvelopment - update
  • Andrew said that Alan has been making commits of his progress towards the COR-text perspective.
  • Andrew noted that Alan was currently using the W3C DOM API, rather than the CellML API proper, to generate the output, because he wants to preserve features in the XML, like comments and the order of elements. This is not provided by the CellML API, because it is irrelevant to the meaning of CellML models. Andre has noted that since order isn't important to the model, it would be better not to preserve it, and this would mean the text view matches the model tree view. Peter noted that it was important that we had a process to reach consensus on important technical decisions. Andrew said that in this case, not everyone has made up their mind about which approach is best - there are arguments for or against either approach.