Bob Gustafson wrote:
> Are the models/metadata/simulation-results in any sort of Version  
> Control System (Subversion comes to mind)?
Models are assigned a 'version' and a 'variant', and the model curation 
workflow means that models are never changed or deleted, instead a copy 
of the model is made in the CellML repository. The URL of the model 
therefore identifies the model.

The file 
describes the difference between a variant and a version (note the 
document is somewhat out of date. In particular, we are using the Zope 
Object Database to store models, not CVS).
> By including VCS tags in the text of the model/metadata/simulation- 
> results, then whenever these files are read/printed, the version  
> number and date can be also displayed. With a version control system,  
> it would also be possible for a user to be inspecting/using one  
> (slightly older) version while a newer version of the simulation is  
> being created.
This is dealt with in the CellML repository, because the URL of the new 
model doesn't become accessible through the site until the workflow for 
adding the model has been completed. The URL is also an easier 
identifier for existing RDF tools to work with, and allows RDF metadata 
to refer to an explicit model.
> Also, the 'history' of a document can be displayed from the vcs- 
> metadata, along with a log of all changes to a document.
The CellML metadata specification provides for the model to describe 
changes which have been made to it in a machine readable format. Please 
refer to
> Simulation results can be declared 'stale' at some time delay or  
> count older than the current version and deleted if storage space is  
> a scarce resource.
Simulation results are not stored in the metadata, only the information 
needed to compute that data (some important metadata still needs to be 
specified, the simulation metadata specification is still a draft and 
not complete yet. Eventually, the simulation metadata should contain 
enough information to reproduce exactly identical simulation results, 
given the model). The idea behind this is that a model and the 
simulation metadata allows people not only to see the data a paper 
author is talking about, but to confirm that it really is the output 
from the simulation, and make modifications to parameters to see what 
effect they have.

Simulation metadata can refer to variables in the model, so it is 
possible that simulation metadata could refer to a variable which no 
longer exists (this would make the simulation metadata invalid). 
However, external simulation metadata would typically refer to a model 
by the complete URL, including the version and variant, and so would 
need to be explicitly updated to point to a new version and variant (at 
which time the author of the simulation metadata should fix any problems 
like this).

Best regards,

cellml-discussion mailing list

Reply via email to