Michael Hucka mhucka at caltech.edu
Wed Dec 7 21:38:37 PST 2005

  rphair> Is this golden rule a part of the L2v1 spec?
  rphair> Remember the golden rule of SBML annotation: You
  rphair> have to record the annotation >you can't
  rphair> understand as such and serve it back when you're
  rphair> done with the model.

In fact, as Nicolas pointed out in his reply, it is not
stipulated in the SBML specification.  Perhaps this should
be considered a fault -- perhaps it should be part of the

It has been a verbal best-practice guideline for some time,
though obviously not articulated well enough.  When an SBML
best-practices document is finally written, the guideline
would belong there.

The closest written thing is currently item 6.11 in the FAQ:

  | 6.11 How much effort should I invest in preserving the
  |      SBML form when round-tripping models through my
  |      software?
  | The first priority should be to support as much of the
  | SBML standard as possible both for reading and
  | writing. You should write using the most interoperable
  | form as possible as described in Question 6.10. To
  | maximize interoperability beyond this requires trying to
  | include as much of an imported SBML model as possible
  | when rewriting it in SBML. This includes preserving
  | annotation data and avoiding mangling id and name
  | fields.  [...]

There may have been past mention of this guideline in
sbml-discuss discussions too, but I didn't troll through
past postings to search them out in writing this reply.

None of this is to imply any blame on you (Robert)
whatsoever for not knowing this.  It's another one of those
things that are just poorly documented.  These are things
that need fixing.


Mike Hucka, Ph.D.   <mhucka at caltech.edu>   tel +1.626.395.8128
Biological Network Modeling Center     http://bnmc.caltech.edu
The Beckman Institute @ The California Institute of Technology

More information about the Sbml-discuss mailing list