Right, I agree it's useful, and if the data model is going to include the notion of subcontexts, then these types of operations should allow a scope specifier. My main worry is in #1 below based on what we've observed in directories. As long as we figure out how to allow the model to be polyarchical, I think we can avoid those problems. If we can't, then we should provide other more generic ways of allowing operations to flow into related contexts.
>>> <gbyrd@xxxxxxxx> 7/5/06 11:00 PM >>>
My reasonaing was that "subcontext" is a particular kind of relationship, one that implied (or at least allows) inclusion. With that sort of relationship, it makes sense to allow a context query to be optionally pushed down into subcontexts.
We also have the more generaly ContextRelationship for non-hierarchical (and non-inclusive) types of relationships.
Certainly worth discussing...
> I noticed the recursive bool in some of the proposed methods in Greg's > list. I have some questions about that: > > 1) Is the notion of parent/child context 'special' enough to warrant > this? I imagine people have all sorts of Context relationships in mind, > and perhaps they even have multiple hierarchies they'd like to view them > in. For example; within a country's government, one way of > hierarchically organizing contexts might be in an organizational manner, > while an alternate view would be in a geographical manner. What we > noticed happening in LDAP/X.500 was that by only allowing a single > hierarchy, enough problems arose that eventually caused most deployments > to flatten their hierarchies and use other forms of relationships. If we > are going to support the notion of hierarchy, we should seriously > consider making it selectable (i.e. recurse = true using the xyz > hierarchy) > > 2) Many of these also take a uniqueID. The wiki indicates that these > ID's are contextually unique, thus it only makes sense that they would > exist in a single context. If one knows the contextually unique ID would > one know the Context for which it's unique, or is the Context not > derivable from the CUID? > > 3) The getSubjects method which takes a filter may prove to be > problematic when recursively searching in the case where schemas differ > among (sub)contexts. This is not true if attribute identifiers are > globaly unique (which IMO would be a good thing). > > Jim > >>>> Greg Byrd <gbyrd@xxxxxxxx> 7/5/06 10:48 AM >>> > > Hi, folks. I'm new to the list -- working with IBM on a temporary > basis. > > Attached is a document that proposes a list of interfaces and methods > for IDAS. This is built upon (and includes) the two proposed data > models that were previously posted to the list. > > There are some open questions, of course. We can use this as a > starting > point for discussion at this week's meeting in Boston. > > > _______________________________________________ > higgins-dev mailing list > higgins-dev@xxxxxxxxxxx > https://dev.eclipse.org/mailman/listinfo/higgins-dev >
_______________________________________________ higgins-dev mailing list higgins-dev@xxxxxxxxxxx https://dev.eclipse.org/mailman/listinfo/higgins-dev
|