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

I think it would be worth investigating how this relates to ideas of Typed 
RDF [1]. That may give a good idea as to when one should do things in the 
space of datatypes, and when more general objects are the better tool.

Programmers have found many advantage to working in typed languages. We find 
the same ideas appear in Category Theory, and the more recent Homotopy Type 
Theory. Types as defined there always come with a notion of identity.
The advantage of this is that it makes it reduces the space to search for 
when considering when two things are equal, and it should also come with a
method for defining equality.

Typed languages are often associated with static compile time typing. That
is perhaps the thing that seems to fit the least with RDF. I have come across
papers that made the case that typing can also be done dynamically though. 

In the case of ”…”^^tpe datatypes we are clearly dealing with things that can be
statically specified, since they refer to values that are functions from a 
fully presented string. And all the well known datatypes come with good notions
of equality. 

Perhaps that is the key criterion for distinguishing when strong typing 
is appropriate and when not.

[1] I pointed to a number of mathematical resources earlier in this very
long thread. https://lists.w3.org/Archives/Public/semantic-web/2020Jul/0135.html
Perhaps this link especially
  https://gitlab.com/web-cats/CG/-/issues/9


> On 24 Jul 2020, at 21:48, Shaw, Ryan <ryanshaw@unc.edu> wrote:
> 
> 
>> On Jul 24, 2020, at 2:43 PM, Hugh Glaser <hugh@glasers.org> wrote:
>> 
>> Also, what about other literals?
>> Presumably, by implying the same approach and implementation structure, we can have:
>> Literals for telephone numbers (ITU compliant),
>> ISO 19160 or whatever it is for addresses.
>> Librarians would presumably like a structured literal for peoples names.
>> Can I have a similar process for URIs as literals, perhaps.
> 
> These all strike me as great and useful ideas.
> 
> What if it were easy to create new datatypes that mapped between 1) DSLs for producing structured string literals (see examples above) and 2) "unpacked" representations as RDF?
> 
> That would also be useful for stuff like:
> 
> * common bibliographic citation patterns as documented in CSL styles
> * Extended Date and Time Format
> * GeoJSON
> * IP addresses
> * all kinds of printed product codes
> 
> - Ryan

Henry Story

https://co-operating.systems
WhatsApp, Signal, Tel: +33 6 38 32 69 84‬ 
Twitter: @bblfish

Received on Friday, 24 July 2020 21:03:25 UTC