- From: Peter F. Patel-Schneider <pfpschneider@gmail.com>
- Date: Thu, 02 Apr 2015 12:46:51 -0700
- To: Arthur Ryman <arthur.ryman@gmail.com>, RDF Data Shapes Working Group <public-data-shapes-wg@w3.org>
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Past experience indicates that specifying the semantics of recursive shapes is not easy. The Shape Expressions W3C Submission is ill-founded on recursive shapes. The axiomatization for shape expressions has counter-intuitive behaviour on recursive shapes that involve negation or exclusive or. Both of these problems can be demonstrated on very simple examples. It is not that recursive shapes with SPARQL are undecidable - I don't see how recursive shapes with SPARQL could be undecidable - it is that coming up with intuitions on what recursive shapes with SPARQL are supposed to mean is difficult for starters and then determining how to formalize these intuitions is even harder. peter On 04/02/2015 11:10 AM, Arthur Ryman wrote: > Richard, > > Recursive shapes are very useful. > > I think there are two subcases: 1. Shapes that use the high-level > language 2. Shapes that use arbitrary SPARQL templates > > For case 1, I am confident that we can precisely specify the semantics, > without being particularly clever. For case 2, in general this is > probably undecidable, however we might be able to say something about > suitably restricted templates > > -- Arthur > > On Sat, Mar 28, 2015 at 4:12 PM, RDF Data Shapes Working Group Issue > Tracker <sysbot+tracker@w3.org> wrote: >> shapes-ISSUE-22 (recursion): Treatment of recursive shape definitions >> [SHACL Spec] >> >> http://www.w3.org/2014/data-shapes/track/issues/22 >> >> Raised by: Richard Cyganiak On product: SHACL Spec >> >> A shape definition may refer to itself, either directly or indirectly. >> For example, in a bug tracker, issues may depend on other issues. Such >> shapes are problematic if there are cycles in the data. Before one can >> determine if a node belongs to the shape, one may first have to know >> whether it belongs to the shape. >> >> Proposals for treating this include: >> >> - Doing something very clever in the Semantics - Syntactic conditions >> that outlaws recursive shape defintions - Allowing recursive shape >> definitions, but doing something to treat them as informative >> documentation only >> >> >> > -----BEGIN PGP SIGNATURE----- Version: GnuPG v2 iQEcBAEBAgAGBQJVHZyrAAoJECjN6+QThfjzeZ4IAJ3r2xMrD+IvTfiHMcc8M2Rj o16YoOG0Zu11SXFGeiE3Q+gSzC50u+CPbNctUOYfjeF/Gcvp/hVgm1dbUQKuCwpa 4vmqbYcbtRXPaZXZGgE9LRENbygFqOTIlnVsl11Y1cQj8Izzj85cnGfpWFA9LEkF G0OOEzi43kUNvq54X3NLcyBIJvlVpvGapsR/MypWHE9F6CzwXHji0PdUw7OFUYQg VxZK+WJTQQbUfT9JAsGgaFJZ63pooi7Hdu4mOVhJcJAfxFO2oQtNdPilo15YGQHk nO2UmPAJ58FF5II1DfVlggw5Nwva10tGuaZ2BjHBaMYt4KwzckQH8mBIKCEj6VM= =mcZ1 -----END PGP SIGNATURE-----
Received on Thursday, 2 April 2015 19:47:27 UTC