we decided the following approach: the webservice will be a way to query the database to find pathways/reactions of interest (e.g.: "show me all reactions/pathways that involve Lactose" or "show me all pathways updated after my last visit on
5/10/2007"). At the end, the user always gets back an ID that points to a pre-generated SBML file that represents the relevant pathway. We may extend this granularity one step further to reactions (so that a custom set of relevant reactions may be embedded in a single SBML-file, regardless of which pathway they belong to).
The webservice has been developed, however is currently under revision to coordinate methods with an alternate Java-based API.
As for the SBML-files: my focus is now on how to integrate our custom annotation in the form of BioPax.