I didn't get a chance to follow up on this when the discussion initially happened, but wanted to mention a few things which are relevant that didn't come up before.
- The basic underlying mechanism for physically partitioned models is the HREF (same as HTML). ArgoUML already knows how to follow HREFs (although it's got a few rough edges). What's needed is support for creating HREFs, managing search lists and/or virtualized references, managing multiple models in the U/I, managing package/model imports, writing multiple XMI files, etc. - This same mechanism is important for profile management as well (a special case of a partitioned model). ArgoUML currently uses a copy-on-reference paradigm for profile elements, retaining no permanent link to the original profile. The profile is an attribute of the editing session, rather than something that is explicitly linked to the project. This is a problem in and of itself. - ArgoUML 0.20 doesn't keep xmi.ids stable across a load/save cycle. This makes any type of configuration management, comparisons, or external management difficult. This problem is fixed in 0.22. - The discussions of various ways to subset the implementation really need to be informed by the knowledge of what's possible with MDR to be useful. A good mental model for the most basic implementation is probably to think about the way HTML pages work. If you delete a referenced element or change its id, you end up with a broken link. Linus asked: > Is [a subset implementation] a small enough step to set up as > a goal for the 0.24 release of ArgoUML? We haven't really discussed timeframes or themes for the next release (whether we call it 0.24 or 1.0), but a basic implementation should certainly be feasible. As I mentioned up top, much of the reading side of things is done already (or supported implicitly by MDR). Tom --------------------------------------------------------------------- To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
