Re: Comments on the ntriples syntax

On Mon, Sep 24, 2001 at 05:00:28PM -0400, Louis Theran wrote:
> As described by, 
> the ntriples syntax seems designed to discourage automated testing. 
> ntriples is  supposed to enable a ``set of machine-processable test 
> cases corresponding to technical issues addressed by the WG.'' If 
> this is really the case, then comments, variable amounts of 
> whitespace and multiple kinds of line separators aren't appropriate, 
> since the best oracle for automated compatibility testing is straight 
> byte comparison (e.g., UNIX cmp).

A problem with straight byte comparison is that it won't 
handle bNodes (aka anonymous nodes). N-Triples does not
define how a processor will generate the name part of a 
bNode.  I expect graph matching programs, such as cwm [1] or
Euler [2], to be useful in determining if two sets of N-Triples 
are equivalent.


However, I think the WG should consider adding a simple graph 
comparison algorithm in future revisions of this WD.

Art Barstow

Received on Tuesday, 25 September 2001 09:13:02 UTC