[dxwg] Add property to link from Distribution -> Dataset (inverse of dcat:distribution)

dr-shorthair has just created a new issue for https://github.com/w3c/dxwg:

== Add property to link from Distribution -> Dataset (inverse of dcat:distribution) ==
Separating out this question (which is distinct from axiomatization of dcat:distribution #120)
DIscussion from #120 pasted below: 

@dr-shorthair commented 22 days ago
Do we need an inverse property here - to allow a distro to point to the dataset that it purports to instantiate?

@dr-shorthair commented 23 hours ago
In this case I think an inverse property makes sense.
A Distro will usually know which Dataset that it is related to.

@akuckartz commented 23 hours ago
No inverse properties please. They are not necessary. But that probably should better be discussed in a separate issue.

@dr-shorthair commented 22 hours ago
@akuckartz : so if I have a description of a distribution (dcat:Distribution) in front of me, how do I know which dataset(s) it is a distribution of? Do I have to make a separate query to the catalogue?

akuckartz commented 15 hours ago
Some information about inverse properties:
https://www.w3.org/wiki/WebSchemas/InverseProperties

@dr-shorthair commented 14 hours ago
@akuckartz - is this ref meant to answer my question above ?#120 (comment) The page appears to be largely about syntax in different platforms.

My point is that
(i) there is a reasonable requirement to be able to traverse from a distro to its dataset, and
(ii) the normal solution for common requirements is a specific property or property path.

Yes, the information might be obtained using additional service calls/queries, but this may not be the optimal solution.

@akuckartz commented 13 hours ago
@dr-shorthair That page contains the result of older discussions about the need for inverse properties. It states at the beginning:

that this should be addressed at the level of syntaxes for Web data, and not by defining inverse properties in the vocabularies

@dr-shorthair commented 13 hours ago
Thanks for clarifying.

I certainly understand that not all inverses are needed, and that they should only be introduced and named where there is a requirement.

However, that is really just a special case of noting that not all potential relationships should be blessed by being given their own name in a vocabulary: whether you call it ontology engineering, information modeling or data modeling, the art is in selecting which edges from the very large set of candidates in a graph we think are important. We do this for efficiency, not because there is no other path to travel that joins the same two resources. Sometimes that will include inverses. But I do not accept a blanket rule to prohibit them - people will encounter different artefacts from different access routes, and it is best if they can see all the relationships that matter (!) without running an additional query.

In this case, there is a whole lot of descriptive information that is more appropriately attached to the Dataset description than the Distribution. I'd prefer to just navigate over to it than have to reason my way over.

@nicholascar commented 13 hours ago
PROV defines some inverse properties (generated/wasGeneratedBy) but not all (used, no wasUsedBy) due to anticipated use. We should do the same.

Not enough people/systems are capable of reasoning or distributed graph assembly to just always declare only the semantic minimum.

@akuckartz commented 3 hours ago
Again: I think that a separate issue should be opened to propose an inverse property.

Please view or discuss this issue at https://github.com/w3c/dxwg/issues/166 using your GitHub account

Received on Thursday, 15 March 2018 20:15:55 UTC