Re: Mechanism, not policy [was: Attribute uniqueness...]

-----Original Message-----
From: David Carlisle <david@dcarlisle.demon.co.uk>
Date: Wednesday, June 07, 2000 4:55 PM
[...]
>In the absolute interpretation the namespace name is for some
>unexplained reason not taken as the supplied string but as
>the absolute URI that it resolves to.

After so many messages it is difficult for me to know where this
reason is "unexplained".  URIs have been around for
a long time, and relative URI references have been found useful
in certian circumstances. A convention, older than XML, is that
an author has the option of using them in the cases in which they
are actually what the author wants.

When a relative URI reference is ued the URI in question is
found by absolutizing the URI reference.  This is web 101.

I have given one specific example of a database application in
which there is one namespace created per data file, and so
giving them very similar URIs is natural and effective and
saves a lot of extra work.

> This means that despite
>the author having specified a relative URI the same resource is
>always located by the namespace name in all contexts. This is
>just bizarre, if that had been the intention then an absolute URI
>could have been used in the first place.


I can't follow that.  "located by"? "namespace name" maybe words
you are using in a way I don't understand.

>Basically the absolute proposal comes from a fundamental
>misunderstanding of the namespace rec:
>That a namespace with name a particular URI _is_ the resource
>identified by that URI.


Yes.  You call it a fundamental misunderstanding -- but you
will have to admit that it is a consisetnt understanding, that a
lot of others have it, and that it is described in the URI specification.
You also have to admit that those who read that the namespace
nsattr is a URI reference and who are familiar with URIs will
naturally come to that conclusion.

>That is just false.
 *That* is just false ;-)

>If it were true then you would need the forbid
>option as relative URI don't refer (in themselves) to resources.

The step of absolutizing is generally considered a short step
so your īn themselves could be seen as splitting hairs.


> If
>the namespace with _was_ the resource with URI equal to the namespace
>name then clearly xmlns="foo" would be undesirable as the namespace is
>(at most) one thing but the relative URI refernce "foo" identifies
>different things depending on context.


Clearly, for the applications yo are considering, then this would
be unwise.  I would therefore strongly advide you to use an absolute
URI.

>But that is not the situation. Namespaces have names which are URI
>references.

Reading this litterally is clearly what you do but not what everyone
would do.  There are two ways of looking at this. Please admit of the other.
It allows the XML specification to use URIs in such a way that the
two technologies can be specified and eveolve indpendently.
I earlier gave an analogy of a towing hitch. All the other vehicles
use a standard towing hitch - they use URI references out of the box.

You seem not to be very familiar with the processing of URIs,
but I would point out that for any application which is at all web-aware,
a URI is a URI is a URI, and it is a great simplification to
just quote the usual URI-reference way of giving one in a document.

There is a great benefit, (I belive) from a lot of applications
using the XML concrete syntax without rediscussing whether <> or []
would be better.   I belive the same is true of applications using
URIs as identifiers and URI-references as the attribute types to quote them.
It is the way things are done.  You can argue for hours or
you can accept the combined wisdom of the millions of people
who have argued before you.

> That does not mean that they are the resources identified
>by those references, anymore than (to reuse an example I used before)
>than machines in nag which are named after english towns _are_ towns.


English town names are not URIs.   If english town names had had the
necessary
properties I would not have had to introduce the URI. See <a
href="Brighton">!</a>

>The namespace name in
>
><x xmlns="http://www.w3.org" />
>
>just is the URI of the W3C home page. That resource doesn't aquire any
>properties of a namespace just because I used it's identifier as a
>namespace name.


That is NOT the case with URIs.  You can propose a totally new system
of identifiers for namespace names if you like , or you can use URIs.
With URIs, if two things are identified by the URIs and those URIs match
character for character, then those things are the same.

I hope this now makes sense.  Even if you feel a strong need to
invent (or to claim the NS spec already invented) a separate system
of identifiers,  I would like you to really earnestly consider using the URI
space.  It is a clean decision - a clean connection between two technologies
core to the web.  Without it, a lot of stuff has to go back to the drawing
board.

>David


Tim BL

Received on Thursday, 8 June 2000 00:20:15 UTC