Re: Blank nodes must DIE! [ was Re: Blank nodes semantics - existential variables?]

On 03.07.2020 07:39, Dan Brickley wrote:
>
>
> On Fri, 3 Jul 2020 at 00:00, Patrick J Hayes <phayes@ihmc.us 
> <mailto:phayes@ihmc.us>> wrote:
>
>     Dieter, greetings.  We are of course aware of this issue.  It is
>     mentioned in the RDF 1.1 overview document in the W3C
>     recommendations,
>     https://www.w3.org/TR/rdf11-concepts/#section-skolemization. But
>     it is not hard to achieve in practice, eg use a pseudorandom
>     number generator seeded with a string of the current date/time and
>     the what threewords location of the generating host, then append
>     the resulting string to a URI  whose path component starts with
>     |/.well-known/genid/|.
>
>
> UUID generation looks to be well specified too,
> https://tools.ietf.org/html/rfc4122
>
> ...though we will too often end up with differing ids for nodes 
> representing the same real world thing.

Yes, you seem to help to lift the record linkage, duplication detection, 
coreference resolution, reference reconciliation, fuzzy match, object 
identification, object consolidation, deduplication, entity clustering, 
approximative match, identity uncertainty, merge/purge, household 
matching, Reference matching, hardening soft databases problem (just to 
mention a few names in a steadily increasing publication flow starting 
at least already in 1948) to a new level. In general, this approach may 
have two consequences:

1) Your URI does not refer to an entity but to a specific description 
and notoriously incomplete description of it. So your space is a nearly 
random space of different ways to refer to sets of objects (that match 
the partial descriptions).

2) As a contributor to our discussion mentioned earlier it then may make 
more sense to describe entities by their property values and viewing the 
Id property as yet another (and mostly randomized noise) property value. 
I am just hoping that the other properties do not have the same type of 
values.

Skolemization was invented an internal (hidden) symbol level trick of 
computational logic and never meant as a proper modelling style for data 
integration on a world wide scale?