Re: [SED-ML-discuss] SED-ML reports

[Lucian S. <lp...@sp...>]

* Nicolas Le novère <le...@eb...> [2010-05-28 08:56] writes:
> On 27/05/10 21:00, Frank T. Bergmann wrote:
> 
> > All I wanted was to produce reports that can be exchanged with other groups!
> > For this there is not enough information there. If that is outside the scope
> > of SED-ML so be it.
> 
> Actually, I really believe it is outside the scope of SED-ML. And I also 
> believe this discussion is an example of many similar discussions that are 
> currently holding up the release.
> 
> SED-ML is not for exchanging reports, but for exchanging *recipes* on how 
> to produce reports. By that, I mean that different tools should understand 
> SED-ML the same way. If SED-ML says which variables to export, the tools 
> should export the same variables. If SED-ML does not say anything about the 
> font or the color, then we do not care about those. The tools can produce 
> the same font or different font.
> 
> Therefore, it is up to us to precise what we cover in SED-ML. But there is 
> arguably not a threshold below which there is not enough information to 
> interpret the report. Because SED-ML is not about the reports themselves.
> 
> Now, producing reports that cannot be interpreted is not tremendously 
> useful. Currently, we only precise what to export and it is true that 
> without labelling the report vectors or ordering them, they are useless. I 
> am deeply against ordering, particularly for things like reports, so I 
> would favor labelling of vectors.

I can understand this sentiment, and I can see that font, color, etc., are 
indeed somewhat beyond the scope of the intent of SEDML.  However, I do 
think that one huge role of SEDML is to allow people to reproduce 
simulation experiments so that they can be compared with the original.  
And if you are producing text files, if you have the same order of 
columns, you make it 1000 times easier to comapre than if you don't.  This 
is what the SBML Test Suite needs in order to do automatic comparisons of 
simulator output against a standard.

I don't see a better place to put the column-ordering information than the 
SEDML file.  I'll definitely agree that this information should not be 
*required*, and that it may be best suited to living in an annotation 
somewhere.  But since one of the primary early clients of SEDML is the 
SBML test suite, it seems reasonable to provide a standard way of giving 
them that information, so that simulators can use it to produce output 
that can be automatically compared to standard output.

-Lucian