- From: Sandro Hawke <sandro@w3.org>
- Date: Tue, 21 May 2013 13:10:35 -0400
- To: Jan Wielemaker <J.Wielemaker@vu.nl>
- CC: Gavin Carothers <gavin@carothers.name>, Andy Seaborne <andy.seaborne@epimorphics.com>, "public-rdf-comments@w3.org" <public-rdf-comments@w3.org>
On 05/17/2013 11:36 AM, Jan Wielemaker wrote: > On 05/17/2013 05:12 PM, Gavin Carothers wrote: >> >> On Fri, May 17, 2013 at 8:02 AM, Jan Wielemaker <J.Wielemaker@vu.nl >> <mailto:J.Wielemaker@vu.nl>> wrote: >> >> I'm sure this will eventually sort itself out as the old versions of >> these formats die away and everybody complies to the latest >> standard. >> That might take a while though :-( Also, nobody says this is the >> last >> revision of RDF serialization syntax. >> >> >> >> The goal of this Turtle standardization effort was to NOT change the >> parsing of any existing (non pathological) Turtle document. If you are >> aware of any changes we made that do change existing Turtle data please >> tell us. All existing Turtle documents should parse to exactly the same >> RDF Graph (with the exception of changes in RDF Concepts 1.1, such as >> plain literals becoming xsd:strings). Parsers need updating to deal with >> interop issues, but documents and data shouldn't. > > The only case I came across with while running my new parser on the > (very) old test cases was (test-28.ttl, #-comment by me). > > <http://example.org/foo> > <http://example.org/bar> > 2.345, > 1, > 1.0, > # 1., (no longer valid) > 1.000000000, > ... > > Whether that is pathological or not is a bit of a border case I'd say. > > Otherwise, I think you are right validity of data wrt. versions. What > remains are two things. Firstly, if the file has a version that is newer > than what my parser supports I'd much rather tell this right away than > generating hard to understand error messages. Secondly, we have a lot of > different formats, most of which produce triples and some quads (with > multiple graphs) and they all look alike. What parser do I take if the > extension/mimetype is lacking/wrong/lost? If that is the most generic > (TriG/quads), I'm beginning to wonder why we have all these other ones > ... My practical advice (and what I expect to do myself) is to use a parser that handles the superset of turtle, trig, n-triples, n-quads, and maybe the obvious subset of SPARQL. (That's what I hoped TriG would be.) For output of graphs I'd use a conservative version of turtle (space before period, not many chars allowed in pnames), and for datasets I'm not sure but I'll probably use N-Quads, unless TriG gets better. Or I'll make up something new and hope it catches on. :) (I'm also thinking about using reification to fit datasets into graphs, but I wouldn't dare mention that in email like this.) Would that work for you, or do you want to be validating -- giving people warnings about data that's not allowed by some spec? -- Sandro > > Cheers --- Jan > > P.s. The most common issue I've come accross with is about > handling %XX in RDF data. I think the standards are > clear, but the daily experience is no fun :-( > >
Received on Tuesday, 21 May 2013 17:10:53 UTC