Re: [rdf12-semantics] Feedback (Part 3) W3C Candidate Recommendation Snapshot (07 April 2026)

Hi Enrico,

many thanks for your elaborative answer. I now understand that requiring 
the IT function to be injective was a deliberate and thought out working 
group decision, and there will even be a WG Note under construction that 
will provide additional explanation. My primary goal was to ensure that 
this wasn't some mistake and also that there will be some way for 
readers of the RDF 1.2 Semantics to understand why the IT function is 
injective. So I accept the decision of the working group.

Best regards
Michael

On 09/06/2026 17:58, Franconi Enrico wrote:
> Hi Michael,
> I am an editor of the RDF semantics document.
> Let me explain why I believe that injectivity is crucial for the correct 
> semantic characterisation of triple terms and propositions.
> You are not the first to object to this choice :-)
> 
>> this is my third (and last planned) feedback for the RDF 1.2 
>> Semantics, which extends my previous feedback by one additional point. 
>> This is a technical (not just editorial or terminological) point and I 
>> needed some time for this to check.
>>
>> * 18) §5, Definition of a simple interpretation, item 6: IT is defined 
>> as an _injective_ mapping from IRxIPxIR into IR. I do not understand 
>> why IT needs to be (always) injective. As far as I can tell this 
>> feature isn't applied or referred to anywhere else in the document, at 
>> least not in any explicit or obvious way. I checked several other 
>> places in the document where the IT function is used or plays some 
>> role implicitly and also checked all the test cases in the RDF 1.2 
>> Semantics test suite that deal with triple terms, but couldn't find 
>> anything were injectivity would be necessary. Demanding that IT must 
>> be injective is a pretty strong requirement, and having such a strong 
>> requirement is IMO not well in line with the other, deliberately weak 
>> and minimalistic requirements of simple entailment. So if there is 
>> really a need for the injectivity of IT (which I may well have 
>> missed), it should be clearly explained in the document. Otherwise I 
>> suggest to drop the requirement, i.e., let IT be just a mapping from 
>> IRxIPxIR into IR. At least for simple interpretations; it might still 
>> turn out to be a necessary feature for other entailment regimes, but 
>> for them it can then still be required as needed.
> 
> The semantics of a logic-based language is given to back and justify the 
> meaningfulness of the language wrt the phenomenon you want to model.
> Here we want to define a resource as _*the*_ _proxy_ of a triple term, 
> so that we can relate this resource to a reifier, or whatever else. So, 
> this resource needs to represent /exactly/ *the* proposition stated by 
> *the* triple term.
> 
>     [I’m writing a document <https://franconi.github.io/ttr-note/
>      > better explaining all this stuff; the document is still in its
>     infancy, but you may read the introduction to get a clearer global
>     picture of triple terms and propositions (and reifiers).]
> 
> So, if this resource has to be in 1-to-1 relation with the triple term, 
> then IT has to be a bijection, hence IT is an injective mapping. Luckily 
> enough, this semantic constraint has no effect impact in simple, RDF, 
> and RDFS entailments, since no (in)equality is expressible.
> On the other hand, this semantic requirement has an impact on extensions 
> having (in)equality, such as, e.g., RDFS+owl:differentFrom or, 
> obviously, OWL.
> Without injectivity, we would get the following unintuitive behaviour:
> 
>     :c owl:differentFrom :c'.
> 
>         does not entail
> 
>     <:a :b :c> owl:differentFrom <:a :b :c'>.
> 
> This is unintuitive since two (semantically) distinct triple terms are 
> meant to denote (semantically) distinct propositions.
> Don’t be confused by the fact that distinct propositions (i.e., distinct 
> triple terms) may possibly be reified in the same way, namely with the 
> same reifier —this happens in the case you have different ways to talk, 
> say, about the same event.
> I hope you will be convinced, and I am open to discussions at any moment.
> 
> As a group, we have already merged your first batch of suggestions (we 
> accepted them all), and we are discussing your second batch of 
> suggestions (Issue #196 <https://github.com/w3c/rdf-semantics/ 
> issues/196>). In the latter case, we are accepting your first proposal, 
> namely we agree on the bad choice of the phrase "extension mapping”; we 
> are changing it to "denotation mapping”. We are discussing about your 
> second point in the same issue.
> 
> Thanks a lot for your contribution,
> cheers
> —e.
> 
> 
>

Received on Sunday, 14 June 2026 15:50:28 UTC