On 09/08/12 21:09, Peter Ansell wrote:
On 10 August 2012 06:02, Andy Seaborne <[email protected]> wrote:
On 09/08/12 11:57, Richard Cyganiak wrote:
Peter,
On 9 Aug 2012, at 00:19, Peter Ansell wrote:
the current Any23 parser completely implements the
N-Triples/N-Quads spec itself including non-standard features such
as unencoded UTF-8 support for both IRIs and literals, relative
URIs, and blank node identifiers that start with numbers (where the
spec says that blank node identifiers must start with a letter).
FWIW, the RDF Working Group (of which I am a member) is currently
overhauling the N-Triples specification, and is likely to define
N-Quads as a W3C Recommendation in this same specification as well.
It is quite likely (but not certain) that this specification will
include UTF-8 support, and will allow blank node IDs starting with
digits.
Support for relative IRIs has been discussed as well, but seems
unlikely.
Best, Richard
Richard's assessment is correct - there is a strand that argues for not
doing N-Quads because TriG is enough but it's a minority. The majority just
keep putting N-Quads on the agenda if it slips off. As a dump format, it is
essential for many people.
I'm on the working group as a representative of the Apache Software
Foundation. For this, and generally, if you have comments or suggestions
just get in touch - while there isn't a official ASF position, I do feel
duty bound to represent everyone. If you think the working group is making
a big mistake, say so. Implementers and users experience counts for quite a
lot.
This does not stop you sending comments yourself via the working group
comments list where the working group would respond. It gets more formal
later in the process when the working group must respond.
Andy
The proposed changes to allow full UTF-8 IRI and UTF-8 Literal support
and arbitrary alphanumeric blank node ids are great. I don't see the
value in relative IRIs though, as there is no possibility of having
@base/xmlns: etc, so it will always have to be completed by the user,
which could lead to difficulties.
That matches Richard's take on the current situation so I doubt that I
need to make a formal comment about it.
It would be great if N-Quads can be pushed through standardisation as
a very simple extension to an N-Triples that also has support for said
UTF-8 and bnode id extensions, to keep the two formats in sync.
Yes, I agree. As a dump format, only absolute IRIs make sense. The
base can come from where data is read, but in the dump case, it's being
moved around and the last thing you want is the data being location
sensitive.
That said, I do use unresolved URIs for simple test data <s> etc as they
also get written into test code but that's a local issue and not a
standards issue.
Andy
PS I should have mentioned - if anyone does want to participate on a W3C
working group, then you can through Apache if you are a committer.
However, if you work for an organisation that is a W3C member, you
should go via that route because they will be the ones having to make
the IP statements.
Cheers,
Peter
