W3C home > Mailing lists > Public > public-declarative-apps@w3.org > February 2019

Web semantics, aka denotational semantics for CRUD operations on RDF datasets

From: Martynas Jusevičius <martynas@atomgraph.com>
Date: Thu, 21 Feb 2019 14:58:37 +0100
Message-ID: <CAE35VmzL3RK-rRJtLBAxYZyr1Oxm9X00p4OA=pQ0utFZCG8ADw@mail.gmail.com>
To: Linked Data Platform Working Group <public-ldp-wg@w3.org>, hypermedia-web@googlegroups.com, public-declarative-apps@w3.org, Semantic Web <semantic-web@w3.org>, public-lod <public-lod@w3.org>
Cc: Tim Berners-Lee <timbl@w3.org>
Hi all,

cross-posting as I think this touches both Semantic Web and Linked
Data and hypermedia.

I finally wrote down something that was bothering me for a while,
namely formal semantics of web interactions:
https://gist.github.com/namedgraph/64ef07b0a3a66092cfdcbf65eefbf00f

The document defines CRUD semantics for 3 different RDF quad-based
protocols: Linked Data, Graph Store [1] and Quad Store [2]. Linked
Data and Graph Store are shown to be orthogonal projections of an RDF
dataset.

The semantics show that URIs built into the RDF data model allow a
formal specification of a uniform, generic web API. I believe it could
be demonstrated that a data model without built-in URIs and a generic
merge operation cannot be used to specify such a uniform API.

These semantics are what Linked Data Templates [3] boil down to, minus
SPARQL. The LDT spec is being updated to better reflect quad-based
semantics.
It is also what I think Linked Data Platform should have specified,
instead of some loose prose definitions. Where LDP falls in this
classification is unclear to me -- it is a form of Graph Store, but
with overloaded and incompatible with GSP semantics.

Note that (POST to) containers does not require special treatment:
membership can be specified as part of the quads payload, and the
resolution of member URIs is a process separate/orthogonal to CRUD.

We discussed this with TimBL and came to a disagreement: what he calls
Linked Data I call Graph Store. However, under the definition of
Linked Data as "[retrieval of] a description of the resource that is
identified by the URI" [4], returning graph contents as a resource
description does not hold in the general case, as graphs are unlikely
to contain triples about themselves.
I think that shows there is conflation between terms "resource
description" and "graph", which I strongly believe are orthogonal
concepts.

Feedback is very welcome.

[1] https://www.w3.org/TR/sparql11-http-rdf-update/
[2] https://lists.w3.org/Archives/Public/public-sparql-dev/2014AprJun/0008.html
[3] https://atomgraph.github.io/Linked-Data-Templates/
[4] http://linkeddatabook.com/editions/1.0/#htoc11

Martynas
atomgraph.com
Received on Thursday, 21 February 2019 13:59:13 UTC

This archive was generated by hypermail 2.3.1 : Thursday, 21 February 2019 13:59:14 UTC