Re: canonical representation

On 01/11/2016 09:53 PM, Gregg Kellogg wrote:
>> On Jan 11, 2016, at 6:10 PM, Dave Longley 
>> <> wrote:
>> On 01/11/2016 04:23 PM, bergi wrote:
>>> Hi all,
>>> one of the issues we currently discuss in the RDFJS 
>>> Representation Task Force [1] is a canonical representation for 
>>> triples/quads and their components. I proposed to use the 
>>> N-Triples representation [2] if there is a definition (the 
>>> specification contains also a interface for variables). One of 
>>> the reasons for me is the RDF Dataset Normalization [3]. It would
>>> be nice, if we could reuse the canonical method for this use
>>> case. But the normalization specification doesn't define the 
>>> serialization. The N-Triples format would allow to escape 
>>> literals in different ways. Therefore a unique way to escape 
>>> literals must be defined.
>> That the specification doesn't define the serialization is just a 
>> temporary failing of the spec. The serialization is N-Quads as 
>> output by the four interoperable implementations:
> Note that the spec defines a “normalized dataset”, which is a 
> restriction on an RDF dataset where all blank nodes are labeled with
>  stable identifiers.

True, but the algorithm itself requires a canonical serialization, so
we need to specify this in the spec -- and it can then be reused as a
canonical concrete serialization of the "normalized dataset".

> There should also be a canonical serialization format defined, which 
> will/should be canonical N-Quads. (Note, that N-Quads doesn’t define 
> this, but N-Triples defines “Canonical N-Triples” [1], which we can 
> use for an internal definition).


Dave Longley
Digital Bazaar, Inc.

Received on Tuesday, 12 January 2016 06:44:00 UTC