no standard file for all the RDF-suggested XSD datatype IRIs

[ektrah]

As specified in rdf-concepts, RDF places special recognition on IRIs of the form http://www.w3.org/2001/XMLSchema#xxx, where xxx is the name of a datatype defined by XML Schema .

By including these IRIs in the set of recognized data types, a graph automatically entails a set of statements about these terms. For example, for any IRI of the form http://www.w3.org/2001/XMLSchema#xxx in D, the graph entails <http://www.w3.org/2001/XMLSchema#xxx> rdf:type rdfs:Datatype (rdfs1).

There is currently no official Turtle file that explicitly contains these statements. Although I guess it's technically not needed, this has not stopped people from inventing their own xsd: vocabulary.

To give a few examples (in no particular order, and not to pick on anyone):

• https://github.com/audacity/audacity/blob/master/au3/lib-src/lv2/lv2/schemas.lv2/xsd.ttl
• https://github.com/semantalytics/stardog-archetypes/blob/master/xsd/schema/xsd.ttl
• https://github.com/w3c/omn/blob/master/omnlib/import/xsd.ttl
• https://github.com/jeswr/on2ts/blob/master/ontology/xsd.ttl
• https://github.com/konigio/konig/blob/master/konig-ontology/src/main/shapes/xsd.ttl
• https://github.com/RDFLib/prez/blob/main/prez/reference_data/annotations/xsd.ttl
• https://github.com/matthieubosquet/ontologies/blob/main/resources/xsd.ttl
• https://github.com/ruby-rdf/rdf/blob/master/etc/xsd.ttl
• ...

So people seem to feel an itch to scratch. (I'm one of them.)

From a "specification usability" point of view, it would be very helpful if there was an official xsd.ttl that contained those statements, documentation, and unambiguously answered questions such as:

• Is xsd:xxx rdfs:subClassOf xsd:yyy if xxx is derived from yyy, e.g., is xsd:integer a subclass of xsd:decimal?
• Is it okay to use IRIs of the form http://www.w3.org/2001/XMLSchema#xxx where xxx is not the name of a data type, such as xsd:maxInclusive?
• Is xsd:anySimpleType a rdfs:Datatype?
• Is it okay to say xsd:xxx a rdfs:Datatype if xsd:xxx is not in D?
• ...

(Not sure if this is the right repository for this issue, but since this repository contains the turtle files for rdf: and rdfs:, it seemed to fit the closest.)

[pfps]

What use would this file be for RDF?

[ektrah]

I think the existence of this variety of xsd.ttl files highlights a gap between what the specification says and what users/implementers reading the specification understand.

So it's not that what the spec says is wrong, but that it doesn't say enough (e.g., about rdfs:subClassOf relationships between XSD types, the usage of other terms defined by XML Schema, etc.).

Having an official xsd.ttl would close this gap and leave no doubt as to which of those xsd.ttl files got it right.

[niklasl]

I agree that this could be of practical value. (I have also defined a "local xsd.ttl" more than once in RDF-driven applications.)

I also think that the XSD IRIs should be served with content negotiation redirecting to RDF representations (i.e. adhere to the "follow your nose" principle). But I suspect that it may be practically hard to do.

[domel]

Would anyone have any objections if I propose such an xsd.ttl file and place it under the ns directory?
We could then decide later whether or not to reference it explicitly in the specification.

Would such a PR be acceptable to everyone?

[niklasl]

I would fully support it. I'd also greatly appreciate if it was served as RDF with conneg support from http://www.w3.org/2001/XMLSchema# (to follow my nose), but I suspect it would take some convincing to achieve.

(As many above, I too have one cobbled together one many years ago for use in RDF namespace completion tooling. The ones proposed above look promising as starters. As noted in RDF Concepts, not all XSD datatypes are suitable for use in RDF.)

[afs]

Would anyone have any objections if I propose such an xsd.ttl file and place it under the ns directory?

rdf.ttl and rdfs.ttl go to dated URIs and have "rdf"/"rdfs" in the name.

SPARQL uses https://www.w3.org/ns/sparql

What would the final URL be?

As noted in RDF Concepts, not all XSD datatypes are suitable for use in RDF.)

xsd.ttl without indication isn't accurate because it is a subset.

Maybe rdf-xsd.ttl.