W3C home > Mailing lists > Public > public-rdf-wg@w3.org > December 2013

Re: Normative reference from Concepts to Semantics

From: Richard Cyganiak <richard@cyganiak.de>
Date: Tue, 17 Dec 2013 09:15:46 +0000
Message-Id: <E8AB6A83-9FAD-4975-8DDC-C17F13DE2451@cyganiak.de>
Cc: Markus Lanthaler <markus.lanthaler@gmx.net>, RDF Working Group WG <public-rdf-wg@w3.org>
To: Pat Hayes <phayes@ihmc.us>

It's simple spec logic. This is all really spec writing 101.

Let's forget about documents for a minute and keep in mind that ultimately, this WG's job is to precisely and normatively define a large bag of terms, concepts, structures, languages, algorithms, relationships, and so on. To deliver those things to the readership, we write them into documents. Due to the large number of these things and due to an expected clustering of audiences, we deliver several separate documents.

Now, many of the definitions rely on other definitions. So there is a dependency graph between the definitions. For example, the definition of RDF graph depends on the definition of triple. The Turtle language depends on RDF graph. The definition of entailment relies on RDF graph (as it is a relationship that holds between RDF graphs).

When a definition relies on another definition in a different document for its full meaning, then there is a normative dependency. One cannot, for example, know if an implementation of the definition is correct without consulting the other document. This is a normative reference.

When a document is mentioned to provide context, background, good practices, implementation advise, and so on, but the host document doesn't rely in its own definitions on terms defined in the mentioned document, then that's an informative reference.

So the distinction between informative and normative references is not one of taste or politics. It is simply a consequence of what is written in the text.

Looking at RDF 1.1 Concepts, it mentions RDF 1.1 Semantics quite a few times, but only in informative material, in particular in the introduction (which is entirely informative) and in a few informative notes. There are no normative definitions in RDF 1.1 Concepts that rely on definitions in Semantics.

Therefore, making Semantics a normative reference would, in fact, be meaningless. If no normative content references it, then what difference would it make? To be a correct implementation of RDF Concepts, another spec has to use the terms defined in Concepts correctly. For example, if a third party syntax spec claims to serialise RDF graphs, but doesn't support blank nodes, then it is wrong, and doesn't conform, because the normative definition of RDF graphs does include blank nodes. Semantics doesn't factor into that.

In summary, whether a reference is informative or normative depends on the text. A reference in informative text can not logically be normative. Whether Semantics is a normative reference or not in Concepts has no effect on conformance to Concepts, and therefore is not even up for debate, unless other changes are made to Concepts that introduce a normative dependency.


> On 17 Dec 2013, at 03:42, Pat Hayes <phayes@ihmc.us> wrote:
> What exactly is the intended implication of having the reference be informative? If this implies that Semantics is not a normative part of the overall spec, then I must formally object to this. As I recall, the 2004 specification documents all cross-referred normatively to one another, as a matter of design. 
> Pat
>> On Dec 16, 2013, at 10:28 AM, Richard Cyganiak <richard@cyganiak.de> wrote:
>>> On 16 Dec 2013, at 16:32, Markus Lanthaler <markus.lanthaler@gmx.net> wrote:
>>> Another question.. currently Concepts normatively references Semantics which
>>> means we have a circular dependency. Is this what we want or shall we
>>> convert the reference to be informative?
>>> Even though it could be argued either way, I think I would prefer to make it
>>> informative. Thoughts (or has this already been discussed)?
>> +1 to an informative reference. RDF Concepts defines the structure, RDF Semantics defines how to interpret the structure. This seems like the correct layering to me.
>> I note that RDF Semantics is only referenced in informative material, so it’s not clear what a normative reference would mean. A normative reference doesn't seem to be necessary to define any of the things that RDF Concepts normatively defines.
>> Best,
>> Richard
> ------------------------------------------------------------
> IHMC                                     (850)434 8903 home
> 40 South Alcaniz St.            (850)202 4416   office
> Pensacola                            (850)202 4440   fax
> FL 32502                              (850)291 0667   mobile (preferred)
> phayes@ihmc.us       http://www.ihmc.us/users/phayes
Received on Tuesday, 17 December 2013 09:16:14 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 17:04:37 UTC