[dsdl-discuss] Re: Namespace renaming

At 2005-12-05 18:43 +0000, Martin Bryan wrote:
>You disagreed with:
> > >b) assigning all elements with a specific prefix to a different prefix,
> > >without changing the URI
>stating that
> > I disagree ... I think (b) is out of scope. Prefixes are cosmetic
> > and irrelevant in an XML instance in general, though not cosmetic and
> > irrelevant in XSLT stylesheets because of the semantics of XSLT
> > creating instances in which prefixes are used.
>
>But the question is whether or not DSRL can be used to transform a document,
>or just to convert its internals for validation purposes. I know the primary
>purpose of DSRL is the second of these options, but must it preclude the
>first? For example, should it be possible to just apply DSRL and then output
>the DOM with these changes in place?

Oh, I totally agree that DSRL should be used to transform a document,
but I do not believe that changing namespace prefixes is a way that
should be allowed.

Since prefixes in the instance are not used for validation (consider
that RELAX-NG use of prefixes is solely in the constraint expression;
as far as I understand one cannot constrain the prefixes used an XML
in any fashion; all validation is done with the URI-qualified names
having already processed the prefixes), then the ability to rename
the prefixes should have a null effect on the validation task. Of
the top I cannot think where a user of DSRL is going to get a net
benefit in validation by having the ability to change prefixes.

As for transforming a document and translating prefixes on the way in
to prefixes on the way out, that is fraught with insoluble
incompleteness ... while simple things can be done simply, there are
holes in how XML is used that would prevent a total solution from
ever working. If a total solution can never work, why implement a
partial solution?

Consider XSLT and the use of prefixed namespace-qualified names in
XPath attributes ... changing the prefixes in the instance and in the
namespace declarations would render the resulting well-formed XML
instance as being invalid to XSLT because the embedded XPath
addresses use prefixes that may not be bound (or worse, improperly
bound) to namespace URI strings.

And even if you could somehow state "we know about XSLT so make sure
XPath addresses are suitably translated" (which I think cannot be
done because how would one rewrite a totally arbitrary XPath address
sitting in an attribute), DSRL would know nothing of my UBL
vocabulary in which I have arbitrary XPath addresses in attributes
and I bind the URI strings through namespace declarations in scope.

In my experience, prefixes should be treated with kid gloves because
messing with them can easily overlook the author's intent.

>You asked:
>
> > Aren't these [cases c and d] accommodated simply by the renaming of
>elements and
> > attributes (thereby satisfying what used to be HyTime architectural
> > forms? I think there is a role for this in UBL and I'm hoping to use
> > DSRL for some experiments).
>
>They probably are, but the problem is that there in no link between the URI
>and a prefix in DSRL.

But we can mandate such a link, couldn't we? Couldn't we state that
whenever a prefix is used in an attribute specifying qualified names
that there must be a namespace in scope binding the URI string?

And, like XPath 1, couldn't we mandate that in DSRL the absence of a
namespace prefix is necessarily indicating "no namespace" and not the
default namespace? I admit I haven't thought this through, but this
is surely my intuition.

>What happens if I want to assign a new prefix to the
>existing name to avoid a name clash?

Not sure how with document-wide context of DSRL (I'm assuming that)
one would be able to do that ... wouldn't the name clash already be
present in the input? I guess I don't understand the scenario you
are positing.

>Do I go xmlns:old-prefix="xyz"
>xmlns:new-prefix="xyz" and then transform old-prefix:x to new-prefix:x.

Well, I think my claim is that we not allow that ... qualified names
in attributes get shortchanged.

>What
>happens to attributes with no namespace qualifier if I do this (do they get
>assigned to the new-prefix or remain assigned to the old-prefix)?

That's easy: XML states that attributes without prefixes are in no
namespace, they aren't in the default namespace and they aren't in
the namespace of the element. They are just in no namespace.

Consider <x:y xmlns:x="urn:x-x" a="a" x:a="xa"/> ... if an
unprefixed attribute were in x's namespace we'd have a collision ...
which we don't have ... so they must be separate attributes ...
therefore an unprefixed attribute is never in the element's namespace
and never in the default namespace.

>You stated:
>
> > I think the URI attributes should be removed and that the URI's be
> > derived from the prefixes and the namespaces in scope.
>
>My problem is that I don't know which prefixes are in scope.

Why not? In XSLT one can examine the namespace axis. In other
languages you can probably either walk the tree (in a DOM-like
situation), or track namespace declarations (in a SAX-like situation)
and successfully keep track of things.

>Do I have to
>declare all source and result element prefixes, or just the result element
>prefixes, in the map?

I don't think you have to "declare" them explicitly (I think this is
a failing of Schematron since it requires namespaces to be explicitly
declared) I think an implementation can simply examine the namespaces
in scope to ensure things are as they should.

>In fact, the map only needs the DSRL prefix, the
>others are superfluous as far as the definition of the map is concerned as
>they only affect the source and result instances! Umm...

True ... I'm only worried about having features in DSRL where we
inadvertently give our users enough rope to hang themselves. Where
something that solves a simple problem simply is too simple to
accommodate real-world situations where DSRL breaks results.

>The problem resolves to:
>
> > I think any result DSRL produces will have to be well-formed XML
> > which will require the namespaces to be properly declared and
> > used.
>
>What constitutes a proper declaration of namespaces in maps that are to
>convert one instance to another instance?

I think there should always be exactly one way to declare namespaces,
that being the XML way of declaring namespaces in scope of
elements. For the processing of any DSRL element, any referenced
prefixes in qualified names would implicitly be related to the
namespaces in scope in the XML instance.

I hope this helps.

. . . . . . . . Ken

--
Upcoming XSLT/XSL-FO hands-on courses:  Denver,CO March 13-17,2006
World-wide on-site corporate, govt. & user group XML/XSL training.
G. Ken Holman                 mailto:gkholman@CraneSoftwrights.com
Crane Softwrights Ltd.          http://www.CraneSoftwrights.com/d/
Box 266, Kars, Ontario CANADA K0A-2E0    +1(613)489-0999 (F:-0995)
Male Cancer Awareness Aug'05  http://www.CraneSoftwrights.com/d/bc
Legal business disclaimers:  http://www.CraneSoftwrights.com/legal
--
DSDL members discussion list
To unsubscribe, please send a message with the
command  "unsubscribe" to dsdl-discuss-request@dsdl.org
(mailto:dsdl-discuss-request@dsdl.org?Subject=unsubscribe)