Re: Turtle implementation report for RDF::Trine

* Gregory Williams <greg@evilfunhouse.com> [2013-04-17 22:50+0900]
> On Apr 17, 2013, at 10:16 PM, David Wood <david@3roundstones.com> wrote:
> 
> > Hi Greg,
> > 
> > Thank you for your comments.  Apologies for not responding earlier.  That was our oversight.  I will also ask the Turtle editors to respond to you.
> > 
> > Alignment with SPARQL syntax was literally the first issue raised for this working group.  Please see the issue and a small portion of the notes and email discussions around it here:
> >  http://www.w3.org/2011/rdf-wg/track/issues/1
> > 
> > The working group did resolve to include this alignment and it was marked "at risk" pending comments from the community.  We will see whether others feel as strongly as you do during the remainder of the comment period.
> > 
> > *Personally* (and thus not as co-chair), I think it is much more important to optimize for user time than implementor time.  Although these changes make the grammar much less clean, they also remove the most common cause of invalid Turtle creation by hand and by code.  That is worth something substantial to many people.
> 
> Dave,
> 
> Thanks for the quick response.
> 
> I took a look at the discussion around ISSUE-1, but didn't see anything directly about the @prefix vs. PREFIX syntax difference.
> 
> I understand you're not an editor, but could you comment on how you think optimizing "for user time" interacts with my concern about having two different rules about trailing dots? My concern is about optimizing for both implementors and users, but since implementations happen once but use of those implementations again and again, I'm actually much more concerned with the potential impact of these grammar rules on users in this case.

I think everyone's gazing into their crystal balls and trying to figure out how to balance these competing constraints:

  simplicity -- primarily, don't confuse authors. secondarily, don't be cruel to developers.

  compatibility with SPARQL -- make it as easy as possible to copy stuff (triples and directives) between Turtle and SPARQL.

  backward compatibility -- there's a *lot* of Turtle out there.

The user time question you raised is exemplified in the case where someone is copying prefixes from a SPARQL query. I would argue that ideally, we'd see one popular representation for prefix (and base) declarations and it would be compatible with SPARQL (fixing the '@'s and '.'s is frustrating for many users). The big question is how reallistic is it that we can migrate there from our current widely-deployed '@' directives and how can we balance short-term and long-term interests.


> thanks,
> .greg
> 
> 

-- 
-ericP

Received on Wednesday, 17 April 2013 14:03:26 UTC