Norman Gray wrote:
....

> The idea of AGAST is that you develop an ontology of your data -- which could
> potentially be based on a formalised version of the model you might have
> already for OAIS purposes, say.  In that, you say things like "a Page is
> partOf a Document", or you create the concepts of PublicRecordsClosure or
> FixedPeriodClosure, which are both types of ClosedRecord, then you merely (!)
> have to ask questions like "is person X provably in the category of
> PeopleWhoCanSeeClosedRecords or something like that.
> 
> The thing I think is important is that, however arcane the term 'ontology' is
> (it does seem to freak folk out), the ontology of your archive probably
> already exists in some whiteboard diagram, or in your head, in such a way
> that it's relatively easy to translate it into an ontology, which (I assert)
> is almost immediately usable as an access control specification.
> 
> The software to do the heavy lifting, there, already exists, and so AGAST was
> just a repurposing project, aiming to show (I think successfully) how you
> wrap this in service layers to make it easy to integrate into a system.
> 
> I need little encouragement to talk more about this!  Does this look like it
> matches the sitations being addressed here?
> 

That certainly looks to me like a proper formalisation of the issues we
were dealing with. It also seems to apply to the other use cases people are
bringing up of restrictions which derive from licensing and IPR considerations.

Thanks for bringing this to my attention - it looks like something we at
the DCC should be drawing attention to.

If my quick reading of it is correct, I think the one area where it doesn't
fit the NDAD model is that where the restriction applies to part of an object.
The X.812 model you refer to implies access decisions are made at the object
level and access is or is not granted to an object. Where the restriction is at
the level of a database cell or a rectangle of pixels on a page, this model
becomes expensive to apply (since you need to treat cells or pixels as objects
and ask for decisions about each of them.) We took the approach that at this
level, the ADF invokes the object access function but passes across information
which an embedded AEF interprets as it grants access to the object, blanking
out pixels, rows, columns, cells, segments of video, etc.