>> For example, models that are designed simply as test-cases >> for particular expected software behaviour - where the structure and >> expected behaviour of the model are well known but not described in any >> specific published article. From what I remember, MIRIAM uses published >> articles as the reference description of a model in regard to MIRIAM >> compliance, but could another source reference description be used in >> model curation, albeit not in conjunction with MIRIAM compliance? > > This is actually incorrect. MIRIAM does not require publication in an > article. It requires a unique reference description, that can be an > article, a technical report, a web-page etc.
ok - thanks for the clarification! > >>> finally that one can reproduce published results *using different >>> software than the one used by the authors*. >> I think this is very important requirement for model curation, but one >> that is quite daunting for CellML models, especially CellML 1.1 based >> models. One question I have is how different different tools need to be >> to satisfy such curation? What I'm getting at is that there could >> potentially be many "different" tools based on a common library (the >> CellML API and code generation service, or libSBML for example), but to >> me these wouldn't be different enough as the actual interpretation of >> the model is essentially done by the common layer underneath whatever >> extra bits the specific tool is adding in. Or is throwing the model >> through a different numerical integrator, for example, enough to satisfy >> this requirement? > > :-) > > We did not formalise that far. You are entirely right of course. Let's say > it very much depends on the model. Some can be simulated in many software > and it is easy to get the results using different solvers. Some tricky > models can only be simulated by a couple of software ... this is also a problem for CellML models since (currently) only applications built using the CellML API C++ implementation from the Auckland group can handle the newer CellML 1.1 models. While you'd hope it never happens, it is possible for errors in the API implementation to show up differently in different applications. Not much we can do about it though, except to be aware of such factors while curating and annotating CellML models. David. _______________________________________________ cellml-discussion mailing list cellml-discussion@cellml.org http://www.cellml.org/mailman/listinfo/cellml-discussion