[wbs] response to 'Triple terms in the subject position'

The following answers have been successfully submitted to 'Triple terms in
the subject position' (RDF-star Working Group) for Niklas Lindström.

> 
> 
> 
> ---------------------------------
> Your preference
> 
> ----
> What is your preference with respect to allowing triple terms in the
> subject position?
> 
> 


 * ( ) allow triple terms in the subject position
 * (x) forbid triple terms in the subject position
 * ( ) no preference either way

Comments: 

In RDF 1.2, a triple will be clearly defined as denoting the logical
proposition itself. This is an abstract entity, and the basic unit of
expression in RDF. When asserted, this proposition holds in the
interpretation. When a triple is used as an object of another triple, it is
a direct reference to such a proposition.

We have worked through many use cases in this working group (all the ones
collected in the RDF UCR and elsewhere), usually involving provenance or
qualification. All of them are solved by describing circumstances
underlying these abstract propositions. Since then, some examples have been
put forward with arguments that the proposition itself needs to be further
described. But none of these examples has stayed with this defined meaning
of propositions as abstract, logical entities.

A triple used as a subject would still invariably denote the abstract
proposition itself. This is crucial. Every suggested use case has indicated
some contextual entity in a domain of discourse, of a type more concrete
than the logical abstraction. These are all examples of when a simplified,
direct proposition is not enough, and entities reifying them are useful. It
does not make sense to conflate the two.

This conceptual error and subsequent problems of using the abstract logical
proposition as a subject was shown already in the seminal example of the CG
report. The reifier design fully addresses this problem, as is being
documented in Concepts, Semantics and the Primer. Both Turtle and SPARQL
1.2 have been adapted to it.


> 
> 
> 
> 
> 
> ---------------------------------
> About allowing them
> 
> ----
> Regardless of your preference expressed above,
>   can you live with RDF-Concepts allowing triple terms in the subject
> position?
> 
> 


 * ( ) yes, I can live with it
 * (x) no, I formally object to allowing triple terms in the subject
position in RDF-Concepts (please develop in the 'rationale' field below).

Rationale: 

There is no practical reason for further extending RDF now to allow
abstract propositions to be directly described as subjects. Arguments for
it are still accompanied by examples where the proposition is typed or
attributed as something circumstantial, further proving how easily this
misconception might be made.

RDF is based on simple, logical propositions of binary predicates. The
subjects of RDF triples have always been names of resources (using IRIs or
bnodes). Literals are not allowed there, nor should nested triples be.

Adding this would force users to needlessly think about the difference
between an abstract proposition and various general or specific
concretizations thereof. Only the latter are called for. Adding triples as
subjects opens up for a proliferation of data in the wild where this choice
is arbitrarily made, leading to accidental appearance of a new class of
conflations. Beyond enabling this conceptual error, this would further add
to the perceived complexity of RDF (multiple and incompatible modelling
choices), and measurable implementation complexity (more code paths to
check). Associated interoperability issues would emerge.

This objection is concerned with practice. If, in the future, concrete
cases and experimental experience emerge, and the rationale for when to
describe the abstraction itself is shown to be easy to understand, adding
support for triples as subjects at a later stage is of course possible
(likewise for literals).


> 
> 
> 
> 
> 
> ---------------------------------
> About forbidding them
> 
> ----
> Regardless of your preference expressed above,
>   can you live with RDF-Concepts forbidding triple terms in the subject
> position?
> 
> 


 * (x) yes, I can live with it
 * ( ) no, I formally object to forbidding triple terms in the subject
position in RDF-Concepts

Rationale: 

This is about the affordances of a data model for encoding and exchanging
knowledge on the web. That includes keeping down the encoding choices to
what is actually called for. That is not the same as restricting
interpretations.

The situation is different for entailment, where there are logical reasons
to consider intrinsic or inferred facts about propositions. But only in the
interpretation, not for being materialized, and certainly not for encoding
in data exchange. The same goes for intrinsic facts about literal values,
such as the number of glyphs in "xyz", or the year of
"2025-02-21"^^xsd:date (or inferred, inverse relations to such values).

For that, symmetric RDF is informatively defined in concepts, as one way of
encoding such entailments (if they need to be materialized as way of
example, or as a convenient form of implementation). This symmetry is also
present in SPARQL 1.2, albeit for other technical reasons. But that is
quite different from encoding RDF data for publication on the web.


> 
> 
> These answers were last modified on 22 February 2025 at 22:38:39 U.T.C.
> by Niklas Lindström
> 
Answers to this questionnaire can be set and changed at
https://www.w3.org/2002/09/wbs/139681/2025-rdf-star-tripleterms-subject/
until 2025-02-26.

 Regards,

 The Automatic WBS Mailer

Received on Saturday, 22 February 2025 22:39:02 UTC