-------- Original Message --------
Subject: RDF WG Resolution Regarding Various Forms of String Literals
Resent-Date: Wed, 15 Jun 2011 16:40:08 +0000
Resent-From: [email protected]
Date: Wed, 15 Jun 2011 12:39:19 -0400
From: David Wood <[email protected]>
To: Manu Sporny <[email protected]>, Ian Horrocks
<[email protected]>, Lee Feigenbaum <[email protected]>
CC: RDF Working Group WG <[email protected]>, W3C SW CG Group
<[email protected]>, [email protected]
Hi all,
The RDF working group resolved our ISSUE-12 [1] today, which is
intended to "reconcile various forms of string literals".
We resolved to accept the proposal at:
http://www.w3.org/2011/rdf-wg/wiki/StringLiterals/AbolishUntaggedPlain
with the modification that preferred output form (SHOULD) is "foo"
> not "foo" ^^xsd:string in RDF; and we recommend that SPARQL and other
> WGs do the same.
Discussion highlighted several possible areas of concern, which we
believe the current proposal addresses. Specifically, it was noted
that:
- The forms "foo" and "foo"^^xsd:string are equivalent input
syntaxes. - The form "foo" is the preferred output syntax. - The WG
suggests retaining the term "plain literal" in documents to avoid
unnecessary rework. Such plain literals would be considered
semantically equivalent to xsd:strings.
NB: This resolution makes *no statement* about language-tagged
literals (e.g. "foo"@en).
We invite discussion regarding the ramifications of this resolution
to other working groups and implementors.
Regards, Dave
[1] http://www.w3.org/2011/rdf-wg/track/issues/12
This resolution by RDF-WG is an attempt to sort out plain literals
without language tag (simple literal in SPARQL terminology) and xsd:strings.
This resolution is subject to it being acceptable to other groups. Touching
datatypes gets the OWL world concerned.
Only xsd:Strings would exist in the abstract syntax.
:x :p "foo" .
:x :p "foo"^^xsd:String .
is a graph of one triple.
The output should be:
:x :p "foo" .
This is obviously a visible change but, in my experience, unlikely to be
much of a problem because data either has xsd:strings or simple
literals, rarely a mixture and if it is a mixture, not on the same
subject/predicate.
Jena memory models are already equate simple literals and xsd:strings
but can return 2 for a match on "foo".
Persistent models see these are different and unrelated.
There are various to implement this. There is no rush - first step is
to see if the resolution actually sticks and is accepted by other
groups, then the community at large.
Then either or both of:
1/ Change Node_Literal/Node(Factory). This could even be turn
xsd:strings to plain literals (the reverse of the resolution) as
minimising the effect on writers.
2/ Get the parsers and writers to sort it out and leave the graph core
unchanged.
I've already been considering a canonicalization phase (e.g. numbers)
after parsing, before inserting into a model. This could be added to
that pipeline.
-------------
Plain literals with language tag are a whole different matter.
One possible outcome is that have a class (not a datatype) of
rdf:LangTagString or some other sensible name.
Andy
