- From: Holger Knublauch <holger@topquadrant.com>
- Date: Mon, 01 Jun 2015 16:00:48 +1000
- To: Data Shapes Working Group <public-data-shapes-wg@w3.org>
Hi Peter, thanks for starting to bring structure into the suggested refactorings (also in your later email). I do believe we need to tackle them one by one, and break them into separate ISSUEs. So below are some references to tickets (I raised a few of them recently): On 5/29/2015 6:39, Peter F. Patel-Schneider wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA256 > > Here is my list of some differences between > https://w3c.github.io/data-shapes/shacl/ (SHACL) and > https://www.w3.org/2014/data-shapes/wiki/Shacl-sparql (SHACL-SPARQL) and my > proposals on how to make improvements as per the resolution of today. > > peter > > > > > The Language > > Top-Level Constructs: The top-level construct in SHACL-SPARQL is uniformly > the combination of a scope and a shape, with other information permissible. > Shapes are uniformly local. Global constraints are different from shapes. > Proposal: Use the uniform SHACL-SPARQL construct. There are naming issues > that then would need to be resolved. On this one I would suggest that you create an ISSUE in which you explain a problem with the current draft and why you believe your own proposal would be better. To me this feels almost like a matter of taste, but I think we need to get to the bottom of this difference. I cannot really speak on your behalf on this topic. > > Scopes: SHACL-SPARQL has more ways of providing scope (although SHACL may > have recently incorporated most of them). > Proposal: Use SHACL-SPARQL scope mechanism. Raised: http://www.w3.org/2014/data-shapes/track/issues/62 > > Classes: SHACL-SPARQL does not do anything special for any SHACL construct > that is also an instance of rdfs:Class. > > Recursion: SHACL permits certain kinds of recursion, but does not correctly > specify how recursion works. SHACL-SPARQL forbids recursion. > Proposal: Forbid recursion, waiting for a proposal on how it can be done > correctly and effectively. Already open: http://www.w3.org/2014/data-shapes/track/issues/22 > > Templates: SHACL has a template mechanism. > Proposal: Keep the template mechanism. > > Functions: SHACL has a function mechanism. > > Non-SPARQL: SHACL permits other execution mechanisms. > > > The Interface > > Invoking SHACL: SHACL has a set of supported operations. SHACL-SPARQL has > one call, with a control graph and a data graph. Note: There is a proposal > for a more flexible design. > > > The Semantics > > Access to control graph: SHACL uses access to the control graph from > constraint to define constructs. SHACL permits access to the control graph > when determining constraint violations. > Proposal; Define the semantics without accessing the control graph. Already open: http://www.w3.org/2014/data-shapes/track/issues/47 > > Extensions to SPARQL: SHACL requires certain extra functions; SHACL-SPARQL > does not. > Proposal: Define the semantics of SHACL without using extra functions. > > Mapping to SPARQL: SHACL and SHACL-SPARQL have different mappings to SPARQL, > mostly because SHACL uses the extra functions. > Approach: If the extra functions are not used, then a mapping more like the > SHACL-SPARQL one is probably needed. Both raised with: https://www.w3.org/2014/data-shapes/track/issues/63 > > RDFS Inference: SHACL does not use RDFS inference. SHACL-SPARQL does. Open: http://www.w3.org/2014/data-shapes/track/issues/1 Holger
Received on Monday, 1 June 2015 06:02:52 UTC