Re: RDF* and conjectures

Am 23. September 2021 14:49:58 MESZ schrieb Fabio Vitali <fabio.vitali@unibo.it>:
>
>> On 23 Sep 2021, at 13:34, James Anderson <anderson.james.1955@gmail.com> wrote:
>> 
>> 
>>> On 2021-09-23, at 12:59:58, Fabio Vitali <fabio.vitali@unibo.it> wrote:
>>> 
>>> ...
>>> 
>>> Now let's come to named graphs. This situation is better from the syntactical point of view, since graph syntax is actually quite reasonable, but worse from the semantic point of view, since there is none accepted. 
>> 
>> is one to understand "none accepted" in the same sense as you described having understood "do not know"?
>
>in its etymological sense, none meaning 'not one': there is no single semantics approved. Zero or eight, both capture the idea behind 'none accepted'. I believe we can agree on this. 
>
>From my (admittedly, limited) perspective, and limiting myself only on what is relevant to me, i.e., the truth value of the content of a graph, it seems that overwhelmingly the concrete attitude is that the content of a named graph is stated just as the content of the outside (the default graph). YMMV. 

i agree, plus a few other things (occurrence, referentially transparent, named by its name) 

>So what we lack is a syntax and a semantics for non-stated graph content.

i agree also that we do lack that (and i'm indeed very sympathetic to your goal!). 

however, we also need a syntax for concise annotation of statements as referentially transparent occurrences, as evidenced by the seminal use case of RDF*, almost all of the use cases this CG collected, RDF* examples in tutorials all over the web and last not least labeled property graphs. 

in your enthusiasm for RDF* you make one essential mistake, just like the proponents of the proposed semantics btw: RDF* is not, as you suggest, a clean slate. to the contrary it has since its conception about seven (give or take) years ago been imagined, understood and advertised with a semantics of referentially transparent occurrences in mind - not explicitly but by virtue of the use cases it targeted. the use cases that drove its popularity determine the semantics of its use today. 

according to the proposed semantics what today is a simple embedded triple becomes two extra statements to define an occurrence plus another statement to define the property in the annotating statement as referential transparency enabling. it is easy to see that this is at best marginally better than standard reification. i would argue it is even worse as it requires a mix of per-statement and per-property-and-graph declarations. 

consequently we will see either
- everybody stopping to use RDF* to annotate statement because the proposed semantics makes it too cumbersome to annotate in semantically correct ways
or
- everybody continuing to use RDF* to annotate statements like they have for years already. 
which one is more likely, what do you think? 

look at the fate of the named graphs semantics by carroll et al. they were published in 2005, everything still in its infancy. they did pretty much what you propose, including a notion of un-accepted statements (very similar to unasserted). yet nobody cared (including the editors of SPARQL) because there was a basic need to group triples that was a much more pressing issue to most.  there is also a basic need to annotate triples, now with the advent of property graphs more than ever. those needs don't bow to some semantics at the very end of some spec. you, and pierre-antoine, can just forget about that. 

semantics are not easy to understand and syntax always has that tendency to look deceptively like exactly the right hammer for your nail. a clever semantics may be able to softly nudge the users into the right direction. it can build dykes, but it can't make water flow upwards. therefor the proposed semantics will fail miserably. but semanticists have a way to not take users serious, maybe in revenge to users not listening to them. but users win every time. 

what could work is to accept that embedded triples are used to refer to referentially transparent occurrences by default - paving the cow paths - and slightly extend the syntax to express referentially opaque types, like eg
<< s p o * >>
with a trailing star with the proposed RDF-star semantics. 
the extra effort in typing and parsing is a very small price to pay to implement an important but still (and forever) niche use case. but the editors would first have to accept that they can't just ignore the use cases. that approach will fail, and rightfully so! 

please excuse my terseness! you certainly deserve a longer answer, but for now this is the gist of it. 


thomas 


>Fabio
>
>
>
>
>--
>
>Fabio Vitali                            Tiger got to hunt, bird got to fly,
>Dept. of Computer Science        Man got to sit and wonder "Why, why, why?'
>Univ. of Bologna  ITALY               Tiger got to sleep, bird got to land,
>phone:  +39 051 2094872              Man got to tell himself he understand.
>e-mail: fabio@cs.unibo.it         Kurt Vonnegut (1922-2007), "Cat's cradle"
>http://vitali.web.cs.unibo.it/
>
>
>
>
>

Received on Thursday, 23 September 2021 15:19:48 UTC