- From: Holger Knublauch <holger@topquadrant.com>
- Date: Mon, 16 Mar 2015 11:32:45 +1000
- To: public-data-shapes-wg@w3.org
On 3/16/15 11:15 AM, Eric Prud'hommeaux wrote: > This is the impasse I'm trying to address with this document. We are > in a position where some folks want an implementation and some just > want a language definition. You are presenting this as an exclusive-or choice. The current SHACL Spec is primarily a language definition, but it also has an a default implementation built-in. How can the latter be a bad thing? I also remember Jose stating that he cannot proceed writing test cases based on the current Spec. But the SPARQL queries serve as a precise definition of the semantics, so I do not see why this is blocking him, nor anyone who wants to provide a non-SPARQL implementation of the SHACL Core Profile. The natural instinct of previous working groups was to create an Abstract Syntax and then some Concrete Syntaxes. But SHACL is a very different language, because it already has its own macro facility built in, allowing the spec to be nicely self-contained. Implementers just need to support a minimum core engine - the rest can be retrieved from the definitions in the SHACL Turtle file. This is a very elegant and extensible design - much better than hard-coding a random sub-set of the language for no convincing reason. Holger > I'm trying to separate documents so we can have both and the language > definition isn't dependent on the implementation. One reason is to > cater to different audiences but a perhaps more pressing reason is > that the langugae definiton can have many interoperable > implementations without waiting for multiple SPIN implementations.
Received on Monday, 16 March 2015 01:33:20 UTC