2011-04-18 During HARMONY
|Editors present:||Frank Bergmann, Michael Hucka, Sarah Keating, Chris Myers, James Schaff, Lucian Smith|
|Editors absent:||(No one absent)|
|Visitors present:||(No visitors)|
|Location:||Rockfeller Research Laboratories, New York City|
Topics about the SBML development process
Are we agreed about the proposed process for voting/selection of L3 packages?
We reviewed a temporary page with formulations of steps for missing parts of the SBML process for proposing SBML Level 3 packages. Our discussions focused largely on the following points:
- Duration for the vote for acceptance of a package should be reduced from the original 6 weeks minimum for a vote, to 2 weeks minimum
- Voting for a package (accept, revise, reject) should not be pure majority, but rather in the following form: (1) to accept, it should be 50% or more of the votes cast; (2) to reject, it should be over 50%; and (3) anything else is revise.
- Packages should be assigned mailing lists. Members of the lists should include at least one Editor. The lists and the working groups should manage themselves; the Editors should not micromanage them.
We also concluded that we need to issue votes immediately for the current packages. In the voting form, we should include the option for people to sign up for the lists right then.
Did we settle version numbering of packages? Is everyone happy?
We had some further discussions about this, and while there was some hemming and hawing, in the end we decided to leave the numbering as described at Community/Wiki/SBML_Level_3_Core/Package_mechanism.
What can or can't a package do?
The basic principle remains that when a package is stripped from an sbml file, the result must still be syntactically valid SBML, thought it may be meaningless as a model). The
required flag on
<sbml> is to be used to indicate whether the behavior of the model is altered by the package.
We were unable to make further statements about the allowable semantics. We could not think of use cases, and did not want to curtail creativity in the future by disallowing things that we didn't know really mattered.
Improve package proposal / package development process
We addressed this with the first discussion item on this page.
Topics about specific L3 packages
Discussion of some questions that arose in developing the 'comp' package
We had separate discussions during HARMONY about the
comp pakage, at more than one point. The result will be that the package stays mostly the same, with some possible simplifications.
Discussion of unresolved issues with 'groups' package
Can different kinds of elements be part of a group, or must they all be the same kind? They can be of different kinds.
Do group identifiers have a numerical meaning? We decided that this would be introduced either in the second version of the groups package OR another package altogether. The initial version of
groups would have no numerical meaning.
What do annotations of the group imply about the members of the group? We settled on the following principle:
- If the annotation is on the listOfMembers, it's a short-hand that means the annotation applies to each individual member as if the annotation were put on the individual members directly.
- If the annotation is on the group, it refers to the group itself.
If there is no assumed meaning for membership (i.e., it's not always is-a), how does a tool know what the writer intended for the relationship to be? Annotations are it, for now.
How to proceed with the layout & render proposals & specs
We need to help rewrite the documents into a proper L3 specification format. No timeline was settled for doing this.
Topics about relationships to other efforts
How can/should SBML interface with SBOL?
We will continue to watch these efforts and see how SBML can interact with them. Crucial to this is the completion of the hierarchy package, since SBOL is largely about instantiating genetic constructs to form genetic circuits. In Chris Myer's work to incorporate SBOL in iBioSim, they will work towards utilizing the hierarchy package to allow for SBML and SBOL to work together.
The creation of the separate SBOL effort may have arisen from a concern about the slow progress on packages, especially the hierarchical model composition package. However, I think that good strides have been made recently, so perhaps SBOL people can be brought around once the
comp package is completed.
Topics about the future of SBML
What to do about parts of the spec that no simulator appears to support?
(Related: Tracker item #2017553 "Can csymbol delay reference another delay?")
No conclusions reached. Topic deferred to another time.
Looking down 2 years or so, how should the chair be replaced?
We discussed various topics but did not develop a process for replacing the chair. The most promising avenue is that Mike be involved in obtaining grants to support the broader COMBINE effort, and/or specific SBML development topics. The time window is 2 years (to 2012).
What to do about SBML Level 4?
We generally agree that it is too early to discuss L4, because if it starts now, people are likely to avoid L3 altogether, hampering SBML's ability to address current needs such as support for hierarchical composition and spatial models.