Re: [xml-dev] xsd versioning

["C. M. Sperberg-McQueen" <cmsmcq@acm.org>]

On 18 Mar 2009, at 04:40 , Andrew Welch wrote:

> Is this considered good practice:
>
> http://www.w3.org/TR/xmlschema11-1/#langids
>
> Seems a bit mad to me.....


I'm not sure whether you're referring to the provision of
a URI to name a language like XSD, or to the idea of providing
several URIs to distinguish different versions of the
language, as well as some URIs to denote the language
itself, without distinguishing versions in all details.

Providing URIs to refer to things seems to me to be good
practice whenever it's desirable to provide a globally unique
identifier for something without having to deal with a
registry (other than the DNS registry).  I don't think this
is particularly controversial, although I don't think everyone
believes it essential.

Providing several URIs to refer to languages and versions of
those languages, with varying degrees of precision, is a little
less common.  I may be biased (I am responsible in part for
the document you refer to), but I think it's good practice.

For any resource X which is subject to change, the question
regularly arises, "when you refer to X, do you mean the version of
X which is current at the time you make the reference, or to
the version of X which is current at the time the reference is
resolved?".

I doubt that there is any simple general answer to the question
"what do you mean by X", but in practice it has proven useful
to provide different ways of referring to X, one with one meaning
and one with the other.  This doesn't make it easier for you to
figure out what you mean -- indeed, it makes it worse, because it
confronts you with the necessity of deciding what you mean, which
is sometimes tiresome -- but if you know what you want to say,
it does make it possible to say it.

As an illustration that the issue is (a) real and (b) not new,
I refer you to the notes that appear in the references sections
of many (but not all) international standards, explaining
whether the references to other standards are to be taken
ONLY as references to the specific dated versions identified,
or whether conforming implementations of the referring spec
are allowed to use later versions of the specs referred to.

The technical reports on the W3C /TR page provide another
example.  These invariably have both dated and undated URIs when
first published.  As later editions or version are published,
the undated version shifts to refer to the newer version, and
the dated URIs continue to refer to the resource as of the
given date.  In my experience, it has been very convenient to
be able to formulate a URI which means "The XML 1.0
spec" (currently, that is, XML 1.0 Fifth Edition, and later
perhaps Sixth Edition etc.), and a different URI which means
"The Fifth Edition of the XML 1.0 spec" (which later, once
a Sixth Edition is published will continue to denote the
Fifth Edition).

For many purposes, the URIs will less precision will be preferable.
But some implementations which now support the current draft of XSD
1.1 for various constructs shipped earlier versions of their
software with support for earlier versions of XSD 1.1.  If they
wish to continue to offer support for those old constructs,
in order to avoid leaving their existing users in the lurch,
it's convenient for them to have reasonably clear ways of
distinguishing different versions of the XSD language.  And so
they asked for ways to refer to earlier versions of XSD 1.1.

Being able to identify things with the desired degree of precision
AND NOT MORE seems to me fairly important in conveying meaning
well.  YMMV, of course.

Do you believe that the rules of RFC 4646 and RFC 4647, and
those of the earlier RFCs on language codes, are mad to make
it possible to give you the choice of saying xml:lang="de"
or xml:lang="de-CH" or xml:lang="de-CH-x-phonebk" or any
of a number of other possible levels of precision?

Michael Sperberg-McQueen


-- 
****************************************************************
* C. M. Sperberg-McQueen, Black Mesa Technologies LLC
* http://www.blackmesatech.com
* http://cmsmcq.com/mib
* http://balisage.net
****************************************************************