- From: Holger Knublauch <holger@topquadrant.com>
- Date: Fri, 10 Apr 2015 09:52:12 +1000
- To: public-data-shapes-wg@w3.org
On 4/10/2015 9:40, Peter F. Patel-Schneider wrote: > -----BEGIN PGP SIGNED MESSAGE----- > Hash: SHA1 > > I consider this approach as extremely dangerous. > > Section 1-6 of http://w3c.github.io/data-shapes/shacl/ forms roughly a > partial guide to a high-level language for shape/constraints. To the extent > that it gives some basic ideas of what might be in SHACL, this document > fragment is not bad, and can serve as a valuable resource for the working > group. However, there are lots of holes in it, ranging from unclear > descriptions of just what is going on in particular cases to missing > information about how checking is supposed to happen to a lack of > fundamental principles. I would like to fix those holes and would appreciate specific feedback. I am aware of the recursion issue with sh:valueShape. Then the textual definitions may need work. But no matter what machine-readable formalism we chose, there is still value in also building a clear textual definition, and my document at least provides a framework to collect and evolve those definitions. > > My view is that the real problem in the working group is that there is no > agreement on what the fundamental principles of SHACL are supposed to be. > Without at least an initial agreement on these principles the working group > is not going to be making significant headway. > > Pushing this document towards a FPWD without some agreement on the > underlying principles of SHACL is only going to further delay the needed > debate on the fundamentals of SHACL, as the working group members who might > be able to work on these principles end up expending effort on things like > the names of particular bits of SHACL or whether the textual description of > some construct is correct. The naming discussions are important because without an agreed syntax we are duplicating parallel efforts and some work (e.g. test cases) cannot even begin. The test cases in turn may help the semantics discussion. Furthermore, the naming discussions are low hanging fruits and the core language should be in a pretty good shape (syntactically) by next week, and we can move on. Holger
Received on Thursday, 9 April 2015 23:53:39 UTC