[Graph] Comparison of two semantics (first step)

As I said at today's telecon, I'd like us to compare Pat's proposed 
semantics to the already existing proposal that is on the wiki [1].

As Andy mentioned in the telecon, the best to do would be that we take a 
look at the use cases and show in what way the semantics (Pat's and [1]) 
help addressing them.

But first, just a high level comparison, I'll send description of how to 
address use cases in a different email:

  - the proposal of [1] is based on Datasets which makes it a bit more 
inline with SPARQL specs, while Pat's is based on quads and quad-graphs. 
In practice, it does not make a big different as you can go from one to 
the other and back without any loss, but it makes the connection with 
SPARQL more explicit
  - the proposal of [1] does not change the RDF semantics or syntax, it 
only refers to it. So a RDF Graph is still a set of triples with the 
same semantics. A dataset is a distinct data structure that is used only 
by people who *want* to separate and manage different graphs, for 
whatever reasons (time variations, provenance issues, trust, 
endorsement, etc). If I understand well Pat's proposal, his goal is to 
make RDF interpretation evolve to the point where they can satisfy 
quadruples as well as triples.
  - since [1] is based on the current RDF semantics, it externalises 
most of the definitions to the RDF Semantics document and the whole 
proposal is very concise. What is in [1] is a complete formal logic and 
can be implement right now.
  - [1] currently assumes that the underlying logic of the individual 
graphs is RDF-entailment, but adapting the definitions to take account 
of RDFS or D-entailment, or OWL, or SWRL would simply need to replace 
the phrase "RDF-model" by "X-model" where X is an entailment regime. 
It's not yet clear what Pat's proposal does wrt RDFS and extensions, see 
espcially Q6 in Pat's FAQ from his email [2].

The proposal in [1] is not really well addressing use cases that are 
important for me, especially UC 6.2 [3] but it could with adaptation. 
Pat's proposal, however, in its main approach (ternary IEXT), would do 
the job very well, provided that some adaptation and clarification are made.

So if I was trying to only address the use cases that are important to 
me, I would have probably gone towards the road of a quad-based 
semantics à la Pat's, but I feel that the Dataset-based semantics is 
better addressing important use cases and would be much easier to 
implement and deploy.

I also think that an educated comparison can only be made if Pat's 
proposal is fully described in terms of formal details.


[1] 
http://www.w3.org/2011/rdf-wg/wiki/TF-Graphs/RDF-Datasets-Proposal#Semantics
[2] http://lists.w3.org/Archives/Public/public-rdf-wg/2012Feb/0094.html
[3] 
http://www.w3.org/2011/rdf-wg/wiki/TF-Graphs-UC#.28B_priority.29_Reasoning_over_annotations
-- 
Antoine Zimmermann
ISCOD / LSTI - Institut Henri Fayol
École Nationale Supérieure des Mines de Saint-Étienne
158 cours Fauriel
42023 Saint-Étienne Cedex 2
France
Tél:+33(0)4 77 42 83 36
Fax:+33(0)4 77 42 66 66
http://zimmer.aprilfoolsreview.com/

Received on Wednesday, 22 February 2012 17:44:26 UTC