Asset class

[dr-shorthair]

Closes #416

Preview https://raw.githack.com/w3c/sdw-sosa-ssn/416-asset-class/ssn/index.html#Systems-and-their-Deployment

[dr-shorthair]

@kjano @KathiSchleidt @sgrellet @afeliachi Can you check that this accomplishes what we discussed.

[sgrellet]

Thanks for this. Two questions from my end

• if we have hasDeployment - deployedAsset. Why keeping inDeployment/deployedOnPlatform ? and systemDeployment/deployedSystem ? for backward compatibility ?
• same with hosts/isHostedBy => a recursive association (hosts/isHostedBy) on Assest with help removing the other two

[maximelefrancois86]

Good proposal, just a few minor change requests

[maximelefrancois86]

UML world... Abstract classes don't exist in RDF and OWL. I suggest removing this

[dr-shorthair]

We already had this form of words for Execution and System.
Yes, this is 'UML-ish' but here it is just normative text.
I don't know how to write an axiom that rdf:type sosa:Asset must not appear.
(It can probably be done in SHACL, of course)

[maximelefrancois86]

OK.
There are two provisions here: (one uses MUST, one uses SHOULD).
The second one can be checked/enforced with ex. SHACL, but the first one cannot (how something is considered is in the head of the user, not in the implementation). So I would suggest to use a different formulation like:

Asset are considered abstract. Entities SHOULD NOT be explicitly typed as
an Asset, unless they are typed with one of the (concrete) sub-classes of Asset.

Maybe this can be part of a separate issue.

[dr-shorthair]

I've changed it to

The Asset class is considered to be abstract. An individual SHOULD NOT be typed as an Asset, but MAY be typed with one of the (concrete) sub-classes of Asset.

Yes, this still has two normative provisions, but in my opinion they belong together.

Also changed Execution and System to match.

[dr-shorthair]

(or should it be MUST NOT rather than SHOULD NOT)

[maximelefrancois86]

I prefer SHOULD NOT.

Many OWL APIs have the option to explicit inferred triples. Saying MUST NOT would imply that we forbid users to use these tools

[dr-shorthair]

Right. The provision applies to assertions, not to inferences. Let's try to make that clear.

[dr-shorthair]

Maybe just

The Asset class is considered to be abstract. Individual assets SHOULD be typed with one of the (concrete) sub-classes of Asset

[sgrellet]

• same with hosts/isHostedBy => a recursive association (hosts/isHostedBy) on Assest with help removing the other two

some after thoughts, with such a high level recursive association, if we still want to keep the hasSubSystem, isSubSystemOf we could make those sub properties of the high level proposed ones

[maximelefrancois86]

OK.
There are two provisions here: (one uses MUST, one uses SHOULD).
The second one can be checked/enforced with ex. SHACL, but the first one cannot (how something is considered is in the head of the user, not in the implementation). So I would suggest to use a different formulation like:

Asset are considered abstract. Entities SHOULD NOT be explicitly typed as
an Asset, unless they are typed with one of the (concrete) sub-classes of Asset.

Maybe this can be part of a separate issue.