RE: [xml-dev] RE: Difference between "normalize" and "canonicalize"?

["Michael Kay" <mike@saxonica.com>]

> 
> Consider the XSLT specification: in Appendix G is the "Schema 
> for XSLT Stylesheets" [1] and it is listed as "non-normative." 
> 
> 1. Why is the Schema non-normative? 

Several reasons.

(1) If a spec says the same thing in two different ways, then it's a good
idea to state which of them takes precedence, in case they turn out to be
different. A convenient way to do that is to make one of them non-normative.

(2) The schema is less complete than the prose as a statement of what
consitutes a valid stylesheet. For example, it explicitly doesn't cover
forwards-compatibility mode (a stylesheet with version="3.0").

(3) It's useful for the reader to know that there is no information in the
schema that he couldn't have found elsewhere.
> 
> 2. What does it mean for the Schema to be non-normative? 

It means it's for information only. You could throw it away, and the spec
might be less helpful, but it wouldn't change the language being described.

Contrast with the schema for schema documents in the XSD specification,
which is normative, and which contains many rules that are not present in
the prose. A significant difficulty in that case has been the presence of
conflicts between the S4S and the prose: for example disagreement as to
whether the lexical representation of xs:positiveInteger allows a leading
plus sign. 
> 
> 3. Does it mean that where there are discrepancies between 
> what the prose says and what the Schema says then the prose wins? 

Yes.
> 
> 4. Why are there discrepancies? When will they be resolved?

Because it was produced by fallible human beings. When they are reported and
fixed, like any other bug. 
> 
> 5. When will the Schema become normative?

It won't. 
> 
> 6. Can someone, outside the W3C, create a normative Schema for XSLT? 

Someone could devise a schema that describes more of the constraints than
the current one, for example by using facilities in XSD 1.1 such as
assertions and conditional type assignment. It might define a subset of the
W3C language (for example, "no forwards compatibility mode"), or it might
define a superset ("XSLT + Saxon extensions"). They could write a
specification of which this schema was a normative part. It would not become
a normative part of the W3C XSLT specification, or of W3C's conformance
rules for the language. However, it's not unknown in the standards world for
third parties to produce specifications that subset or superset a base
specification where there is a need for it in the user community.

> 
> 7. Or, is creating a Schema for XSLT exclusively within the 
> jurisdiction of the W3C XSLT working group?

The W3C does not have any jurisdiction. It publishes specifications that the
community can accept or ignore as it pleases. Anyone else can do the same.

W3C could choose to use intellectual property mechanisms to prevent other
people producing variants of its specifications. I don't know what its
policy is on this.

Michael Kay
http://www.saxonica.com/