Single vs. Collection of Asset and Party unclear

[nitmws]

The Information Model defines the Asset Class this way: An Asset is a resource or a collection of resources that are the subject of a Rule. ... and the subclass AssetCollection that way: an Asset that is a group of resources. This indicates that all the items in the group will be the subject of the Rule.

Unclear is:

• Wording: why are "some assets" called "collection of resources" and "group of resources" - if the same is meant in both cases. Let's use the same term.
• Design: how to indicate that an Asset is only a single resource and not a collection? Using the class Asset covers both, the class AssetCollection only collections - nothing specific for a single resource.
• Check: this design allows nesting collections in a collection as both the subject and object of an RDF triple with odrl:partOf as predicate may be an AssetCollection. This could result in deeeep nesting - ok?

These issues above apply to PartyCollection too.

[riannella]

1) s/group/collection/

2) There is not definitive way to indicate a (single) Asset - as we are only concerned with Collections to ensure the constraints are applied correctly

3) We have no restrictions on levels of partOf

commit: 9bfdc6fae0d75cc4be9679b834afefa7b9b723b0

[simonstey]

Design: how to indicate that an Asset is only a single resource and not a collection? Using the class Asset covers both, the class AssetCollection only collections - nothing specific for a single resource.

An asset can be a single asset as long as it's:

1. not of type AssetCollection, and
2. has no assets that are part of it.

[nitmws]

@simonstey the tripwire of 2. is: it is not required that a used AssetCollection must list all of its "parts" in a policy. Therefore at the time of processing a Policy the processor does not know exactly if this asset thing thing has other Assets claiming they are a partOf it.

[riannella]

Yes, that is correct. We can't really do anything about that...can we?

[simonstey]

@simonstey the tripwire of 2. is: it is not required that a used AssetCollection must list all of its "parts" in a policy. Therefore at the time of processing a Policy the processor does not know exactly if this asset thing thing has other Assets claiming they are a partOf it.

that's why we shouldn't adopt the Open World Assumption for evaluating/processing policies. otherwise, you also wouldn't know whether a policy is really linked to all of its rules, a rule is really listing all of its assignees, assigners, etc.

Yes, that is correct. We can't really do anything about that...can we?

sticking to CWA for policy evaluation, i.e. everything that's not explicitly stated does not exist.

---

fwiw, following SHACL shape can be used to check whether an asset conforms to the requirements of single assets (i.e., not of type AssetCollection and not partOf another Asset)

odrl:SingleAssetShape
    a sh:NodeShape ;
    sh:targetNode ex:partB ; # replace ex:partB with the asset you want to check
    sh:not [ sh:class odrl:AssetCollection ] ;
    sh:property [
    sh:path odrl:partOf ;
    sh:maxCount 0 ;
    ] .

[riannella]

We only know what we know ;-)

Perhaps some of this more semantic processing we document in the Formal Semantics Note?