- From: Bert Van Nuffelen via GitHub <sysbot+gh@w3.org>
- Date: Mon, 22 Mar 2021 08:22:51 +0000
- To: public-dxwg-wg@w3.org
Conformance is very complicated story. In particular if one takes into account machine conformance checking. A first observation that one has to make is what should be included in the conformance check. That is open for interpretation. Is the usage of the label part of it? Is the usage of the definition part of it? Is the usage of the Usage Note part of it? The proposed conformance statement on relying on RDF only guarantees that the URIs are being used according to the proposed structure. So if one leaves the above open (which is a fine choice), then one user will insist on implementing the usage note https://www.w3.org/TR/vocab-dcat-2/#Property:distribution_media_type while the other ignores it as it is just a usage note. Conformance is thus about making choices, and the more we aim for (technical and semantical) harmonisation, the less freedom is possible. Similar argumentations happen with SHACL. SHACL seems to be a gift to bring clarity in this discussion. A machine will tell if my RDF serialization is conform the specification. Unfortunately itself brings new options to the table, called inference. And because (limited set of) inference is inherent part of the SHACL language it is even more tricky. In DCAT-AP community there was a long discussion on whether the range constraint of being a member of a class (E.g. https://www.w3.org/TR/vocab-dcat-2/#Property:resource_theme: the range of dcat:theme is skos:Concept) should be part of the supplied data to be DCAT-AP conform. So should it be ``` _:dataset1 dcat:theme nal:EDUC. ``` or should it be ``` _:dataset1 dcat:theme nal:EDUC. nal:EDUC a skos:Concept. ``` One can argue for either of both options. See https://github.com/SEMICeu/DCAT-AP/blob/2.1.0-draft/releases/2.1.0/examples/examples.md and https://www.itb.ec.europa.eu/shacl/dcat-ap/upload for more choices one has to consider. For me the SHACL conformance challenge is problematic because it seems that one has a formal language but that one still can get to two different conformance outcomes. But it is has one benefit, because one can technically inspect the used constraints and inference. So the difference or equivalence can be technically explained. (But the means have to be given to the users.) The first conformance is much more complicated, and is about the nature of the semantics and harmonisation one likes to achieve across all profiles. -- GitHub Notification of comment by bertvannuffelen Please view or discuss this issue at https://github.com/w3c/dxwg/issues/1332#issuecomment-803864006 using your GitHub account -- Sent via github-notify-ml as configured in https://github.com/w3c/github-notify-ml-config
Received on Monday, 22 March 2021 08:22:52 UTC