Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
RE: [higgins-dev] Revised Higgins data model goals

I’ve captured Jim and Raj’s comments in additions to [10] here: http://spwiki.editme.com/RevisedDataModelGoalsM4

 

Jim wrote:

 

Yeah, the org example points at the need is where (as Paul pointed out as well), some kind of relationship needs to exist and it may not make sense to draw it at the DS level.

 

Jim

>>> Nataraj Nagaratnam <natarajn@xxxxxxxxxx> 4/18/06 9:44 PM >>>

Yes, context relationships need to be managed - whether they are hierarchical or not depends on the relationship, and importantly, like Jim was pointing out, depends on the semantics of the consuming application and applicable policies.

Thinking of an example - would we characterize Higgins as a sub-context to Eclipse in org sense? Does that mean all policies applicable ot Eclipse are applicable to Higgins? Does it apply to membership? access list? Probably not everything. So, there are relationships.. and information about relationships that may identify some hierarchy in organizational structure, geo affinity (NC, UT, MA are states in USA), etc.  

Raj




"Paul Trevithick" <paul@xxxxxxxxxxxxx>
Sent by: higgins-dev-bounces@xxxxxxxxxxx

04/18/2006 09:14 PM

Please respond to
"Higgins (Trust Framework) Project developer discussions" <higgins-dev@xxxxxxxxxxx>

To

"'Higgins (Trust Framework) Project developer discussions'" <higgins-dev@xxxxxxxxxxx>

cc

 

Subject

RE: [higgins-dev] Revised Higgins data model goals

 

 

 





Jim wrote:
>
> Oh, now I see two discussion points (different types of relationships, and
> the notion of hierarchy):
>
> In your description of Mary's example, you're talking about context
> relationships as they are implicitly formulated via references on a
> DigitalSubject.

Yes.

> Not an explicit Context to Context relationship.

Right.

> On the
> other hand, you say Raj and Tony want explicit relationships.

Raj and Tony want the option to be able to express explicit Context to
Context relationships that may or may not be hierarchical.

> These are
> (in my mind) two different kinds of context relationships. (Q1) In the
> data model, are they seen as being lumped together in the same set?

The Raj/Tony kind are supposed to be captured with [10]: Contexts are
associated with other Contexts in 1:1 or 1:M relationships.

The other kind isn't explicitly stated anywhere. Perhaps we need to add back
in an explicit separate goal related to the ability to unidirectionally
correlate DigitalSubjects across contexts and to the "implied" Context to
sub-Context relationships that are thereby implied by the DS-to-DS
unidirectional relationship?

>
> In both of these you mention hierarchy where the target is subordinate
> (maybe you didn't mean to say "hierarchy" in #2 below when describing Raj
> and Tony's case). (Q2)Anyway, what is the benefit of using hierarchical
> terms (subordinate, superior, parent, child) to discuss these
> relationships? If there is none, my fear is people will begin to
> imagine/assume hierarchies where there are none, or be confused that
> hierarchies appear to change (any hierarchy is only valid from the
> perspective of a given DigitalSubject * as if you grabbed a node of a
> graph and lifted it up until all the attached nodes dropped into a
> hierarchy).

Since DS-to-DS relationship links across Context are uni-directional then
there is a "directedness" in the relationships between their respective
containing Contexts. This *is* a kind of hierarchy, though an implicit one.

The Raj/Tony explicit Context-to-Context 1:1 or 1:M may or may not be
hierarchical.

>
> I guess what I'm saying is, this kind of hierarchy can be imposed and
> viewed by a consuming application without us using hierarchical terms in
> our model, and it might be safer that way. By safer, I mean this: In X.500
> / LDAP, the data model has the notion of hierarchy. This is exploited to
> overlay different types of semantics (i.e. A policy applied to one node is
> assumed to be applied to all nodes below it). Overlaying semantics works
> great when everyone agrees on the hierarchy, but in reality, hierarchy,
> and its myriad of implied semantics becomes a political nightmare. So much
> so, that people often shy away from using hierarchy for anything, and
> instead, drawing their own semantic-specific relationships (which is what
> I assumed this data model was striving to allow).
>
> Sorry, I'm rambling about this hierarchy stuff, it's not (nor are any
> hierarchical terms) even mentioned in the revised goals.

I'd be interested in what Tony and Raj think about what you're saying here.
Maybe this is what was behind them stating that the Context-to-Context
relationships are definitely not necessarily hierarchical. Perhaps they were
thinking they should be more loosely associated and then let, as you say,
the apps/policies fight it out at their level.

>
> The expounding I was looking for was an example of where an application
> (or user) would need to acce ss the context relationships. I'm just trying
> to understand the justification for that aspect of the model. With my new
> understanding, it would also be good to clarify Q1 above as well.

Okay.

>
> Jim
>
>
> >>> paul@xxxxxxxxxxxxxxxxx 4/18/06 6:12 PM >>>
> 1) Mary describes a *hierarchical* kind of 1:1 or 1:M relationship. When I
> was discussing this point with Raj and Tony they also envisioned
> non-hierarchical 1:1 and 1:M relationships.
>
> 2) There are two kinds of hierarchies in Mary's example. A DigitalSubject
> in
> one context might have (unidirectional) references to a DigitalSubject in
> another context. From this subject's perspective the latter is a "sub"
> context of the former. This is a kind of implicit Context hierarchy. The
> other, which is what Raj and Tony had in mind in 1) above is an explicit
> Context hierarchy (never mind whether there are any DS to DS relationships
> between them.
>
> Jim: I'm not sure what kind of expounding would be helpful. Let me know.
>
> Mary wrote:
> >
> > A context can have a subcontext (1:1) or multiple subcontexts (1:M)
> >
> > For example, there might be a context associated with a particular
> > system, and a person may be participating with that system in multiple
> > subgroups:  a My Base Ball Fan Yahoo group and My Neighborhood
> > Association Yahoo group.
> >
> > Or a context created for participation at a conference may be structured
> > with subcontexts: one of which corresponds to the person's role as an
> > attendee, one their role as a presenter, and one designed for any
> > visitor to the conference city.
> >
> > >Can we expound on [8] Contexts are associated with other Context s in
> > 1:1
> > >or 1:M relationships?
> >
> > >I know there may be objects in a context which show relationships to
> > >objects in other contexts, and that's one way we might have context to
> > >context relationships, but this goal (I think) is talking about direct
> > >relationships.
> >
> > >What it looks like is someone will want to, given a context, see what
> > >other contexts are related. If so, what will they do with that
> > >information? One case may be where the Contexts are actually replicas
> > of
> > >one another. In that case, the consumer of the context relationship
> > >information is storing failover information. Probably not the use case
> > >that drove the goal.
> >
> > Jim
> >
> >
> >
> > _____________________________________________ __
> > 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
> _______________________________________________
> 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


Back to the top