Re: Request for feedback about RDF-star

On Wed, 24 Mar 2021 at 09:25, Pierre-Antoine Champin <> wrote:

> Dear Semantic Web community,
> as you may already know, an informal "task force" has been created in the
> RDF-DEV Community group [1], in order to produce a specification document
> for RDF-star (née RDF*) [2]. RDF-star extends RDF with native support for
> talking about RDF statements (as an alternative to standard reification),
> and already has a number of implementations. The goal of this work is to
> help ensure that all implementations are actually interoperable (which is
> not quite the case at the moment). Once this specification reaches a stable
> state, and provided that we get enough interest from RDF implementers and
> users, we will try to push it to Recommendation track.
> We require your feedback on the following question. We aim to mint a new
> IRI to be used with RDF-star. In your opinion, is it acceptable/desirable
> to propose its inclusion in the RDF namespace [3], or should we instead
> mint it in a separate namespace? We could not really reach consensus in the
> group, hence we wish to get more opinions from the larger community.
> We understand that, as a Community Group, we have no authority to
> *actually* update the namespace (this would be done only by a future WG).
> But if we succeed in bringing this to a REC, changing from rdf-star:X to
> rdf:X at that point will be impossible in practice (remember the "0.1" part
> in FOAF IRIs?). And we also want to avoid repeating the confusing namespace
> dichotomy of rdf: and rdfs:. If we don't make it to a REC, this will mean
> that RDF-star is not widely used anyway, and so our "polluting" the RDF
> namespace will have had no real impact.
> Some people in the group, on the other hand, feel that the RDF namespace
> should considered fixed (although other specs have already added terms to
> it [4,5]), or that the semantics of RDF-star is not stable enough.
> The whole discussion can be found in the minutes of our call [6].
> Thanks in advance for your feedback.
My advice would be to do something new. Perhaps we could arrange for a new
short memorable ns at W3C that could be used for this?

It isn't clear at this juncture whether RDF-star is the seeds of the next
generation of RDF, or a useful exploration. There are other approaches in
the broad area (e.g. Wikidata's data model, labelled Property Graphs) and
it is quite possible a future REC-track group might take another approach.
So presuming upon official inclusion into the main RDF namespace seems a
little presumptive of us, even if the hope is that things head in that
direction. It could also feel like unfair on the W3C team to have us say
"hey, millions of documents think that "foo" is in the rdf: namespace, how
about making that true?".

Maybe there is something that could be said in the implementation guide for
software-creators to encourage it to be possible/easy to accept a future
term that *is* in W3C's RDF ns?

It's always been an awkward namespace btw, and is one of the oldest XML
namespaces (the XML Namespace technology was designed at the same time, and
not without controversies). One reason it is awkward is that it contains
symbols that are used purely for the RDF/XML syntax designed back in 1997
(rdf:Description), but also it is just horribly long and hard to remember.

>   pa
> [1]
> [2]
> [3]
> [4]
> [5]
> [6]

Received on Friday, 26 March 2021 09:04:13 UTC