ABI CellML Meeting Minutes, 13th July, 2011

Present: Dougal Cowan, Caton Little, Andrew Miller, David "Andre" Nickerson, Tommy Yu, Poul Nielsen, Randall Britten

 

This week's agenda:

 

1) Update on the status of CellML API binaries (Alan Garny)

Discussion:

  • Andrew: have emailed Alan to let him know that once CMake is up and running, the binaries will be provided.

2) CellML API scope discussion - for example, should the CellML API include SED-ML services?

Discussion:

  • Andrew: the name of the CellML API leads to expectations about the scope - should we re-name it to better reflect the scope? Another alternative is to separate out the parts, but this causes problems for users to obtain and use all the parts.
  • Andre: the API should deal with CellML only. Even the integration service should be separated out, as this is a different task from processing CellML. SED-ML support is separate and independent of model encoding.
  • There was a discussion about various services in the API and whether they could or should be separated.
  • Poul: it might make sense to separate current services into modules, eg. CCGS, CIS, etc.
  • Andrew: services are currently separated into directories in the source tree. To further separate them we would need some additional infrastructure.
  • There was and extensive discussion about how to decide how to decide what services go into the API.
  • Randall: suggest we communicate to the community that we will separate the services (such as SED-ML services) when it becomes appropriate. These do need to be separated cleanly, but now is not the right time.
  • Andre: we need to use standard [widely used de-facto standard] libs instead of writing more code that we then need to maintain. We should be trying to use the same libs as libsbml. We still need improved documentation to make what we have now easier to use.
  • There was a discussion about decisions to write our own libs instead of using existing ones.
  • Randall: it may sometimes make sense to compromise on formal standards compliance in favour of de-facto standards in order to enhance familiarity and usability for users. Attracting more developers to use the API would be an advantage, as was reiterated in the survey comments.
  • There was a discussion about options such as rationalising the API, re-writing in another language such as Java, using de-facto standard libraries, etc.
  • Andrew: the current documentation should probably be more readily available on the website. We could collect the documentation together and then improve it after the binary release.
  • Randall: we should have a discussion about alternative names for new and existing services, eg. code generator instead of CCGS.

3) CellML API - update

Progress:

  • Andrew: CMake now automatically generates Java bootstrap code.
  • Andrew: Code tree cleaned up (on branch) to remove autotools files that are no longer needed. 
  • Andrew: CMake build now seems to work fine on Linux, using the standard cmake macros to detect the location of tools like Java, bison, and flex.
           - next step is to get it going on all platforms. 
  • Andrew: CCGS can now generate code for:
          * Picking a realisation at random from a set of realisations.
          * Calling a function to sample from a probability density function. 
  • Andrew: CIS now can use that code, and includes an implementation to numerically sample from a distribution specified using a probability density function. 
  • Andrew: Started looking into how we can represent sampling sensitivity analyses in SED-ML; came up with a basic proposal, and have commenced work to support that proposal in SProS and SRuS. 
  • Andrew: SProS and SRuS now keep track of where a SED-ML document was loaded from and use it when loading models, so relative URLs are handled correctly. 
  • Andrew: I've looked into the SED-ML specification issue around resolving XPath namespace URIs; XPath doesn't explicitly say where namespace URIs are resolved from (it is up to the host language), but other W3C host specifications for XPath resolve relative to the host document, not the target. The SED-ML specification doesn't mention how to handle XPath at all - so I have reopened the existing tracker artifact so this issue isn't forgotten. 
  • Andrew: Some correspondence with Alan about timeline for getting builds ready.

Discussion:

  • Poul: what is Caton's role in API development? Randall: yet to be determined.
  • Andre: documentation should be a top priority.
  • Randall: Peter wants binaries to enable Alan to work on OpenCell to be first priority.
  • Andre: is the brackets issue raised by Lucian fixed yet? Andrew: not yet, some work has been done.

4) CellML core specs - update

  • A breakaway was called.

5) Repository contributions - update

  • No update.

6) PMR2 development - update

  • Tommy: worked on the Z3C form security issue.
  • There was a discussion about security issues with PMR2 webservices.