I agree with Ralph and Frank. Conversion should be a done independently
whether the original or resulting model is valid. The validity checks
should in general not be anything libsbml does automatically they
should be specifically requested by the user. The validity of a
SBML document should be the responsibility of the user. Therefore
libsbml should help to create valid documents but it should no prevent
the user from creating invalid ones.
A more complicated question is whether lossy conversion should this be
allowed. There are a lot of examples where we have this situation
already and both documents are actually valid however the resulting
interpreted model will be different.
The loss of information should be brought to the attention of the user
but I still would think that the conversion should be done. After all
we do not know what the user wants to do with the model and the loss
may be irrelevant.
Stefan Hoops, Ph.D.
Senior Project Associate
Virginia Bioinformatics Institute - 0477
Bioinformatics Facility II
Blacksburg, Va 24061, USA