Re: namespace usage as assertions

At 12:08 PM -0500 6/21/00, Dan Connolly wrote:

>In particular, if the user chose a namespace name beginning
>with http:, you can infer that a description of/representation of
>the resource identified by that URI is generally available on demand.


You _can_ infer that, but not based on any assertions made in the 
namespaces recommendation, which explicitly exempts applications from 
having to make that implication.


>(I say generally because this is a distributed system,
>and availability isn't guaranteed)
>
>It's a matter of dispute whether the resource identified
>by the URI is the same thing as the namespace identified
>by the URI, but how HTTP URIs work is standardized.

yes, but what is retrieved by an http URI cannot be standardized. 
What _can_ be retrieved is some entity-body with some MIME-type. Even 
the assertion that a URI identifies a single resource, which may come 
in several formats is operationally vacuous. Consider an http: URI 
that (depending on content-type request) produced a GIF file of my 
dog, My resume in PDF, a jpeg of my daughter, an HTML file discussing 
"Buffy the Vampire Slayer", and a flash movie of baboons. This may 
seem to violate the intent of URIs, but I'm certainly philosophically 
able to claim that this is an abstract object representing a key 
slice of my life, and that each of these items is the best way to 
describe this object in that medium.

The issue of whether the "thing" at the end of a namespace URI "is" 
the namespace, or "is" a description of the namespace is totally 
vacuous without a definition of conditions that the abstract 
resources must satisfy, *that are also* backed up with particular 
MIME-types and conventions.

Otherwise, whatever your opinion may be, it has no reflex in the code 
that can be written, and thus doesn't matter.

I recently heard a reading from a  hypertext novel called the 
"unknown" (search for it in google). For artistic reasons, and 
because they have the technical chops to appreciate a joke like this, 
I could easily see them creating a namespace whose URI, interpreted 
as a definition, in the way that you advocate, would return a 404 
error.

For abstract objects that are not tightly constrained, anything could 
be a "description" if someone says that it is.

>  > In other words, since there's not necessarily anything "at" the URI that
>>  defines the set of names in the namespace independently of what is
>>  stated in documents that might refer to the URI (documents which might
>>  be created by many different people), there's no way to tell if a given
>>  use is actually correct in asserting that a name is part of a
>>  namespace.  For example, if userA says
>>
>>  xmlns:x="http://foo"
>>
>>  and then says
>>
>>  <x:Title>
>>
>>  there isn't necessarily a way to independently verify that there really
>>  is such a namespace that includes the name "Title".  All you may be able
>>  to know is that the user has made that assertion.
>
>True, there isn't necessarily a way. But I expect that in many
>cases, there will be a way, and I think that cases where there is
>no way should be avoided.

So avoid them. It's unenforceable in any case, and as a philosophical 
argument, it's not very tight. There's nothing to prevent people with 
different philosophies about namespaces to ignore your desired 
practice, while yet claiming to satisfy the letter of any standard 
you might hope to propagate.


>For example, if userB writes
>
>	<html xmlns="http://www.w3.org/1999/xhtml">
>	<banana/>
>	</html>
>
>then userC can, independently, access http://www.w3.org/1999/xhtml
>to get an HTML document that says that the use of that namespace
>name is governed by the XHTML 1.0 specification, look in that
>specification, and discover that there is no 'banana' element
>in the XHTML namespace at this time, and that every html element
>must contain a head and a body.

That's nice, but it's not mandatory, and can't be made mandatory in a 
meaningful way without establishing some language for those 
descriptions. That's possible, but it's hard work, as we've seen. You 
could do worse than implement the ideas that Rick Jeliffe has 
proposed.

In the meantime, uniquely identifying elements requires no such 
assumptions about retrieving data, and should not be held hostage to 
someone's theory that such retrieval is going to be a great idea.


>In due course, an XML schema may become available from that URI,
>at which point the discovery that this document conflicts with
>the namespace name issuer's intended usage can be automated.

That would be cool too, but it's not a prerequisite for recognizing 
that an html:banana is different from an rdf:banana (assuming 
appropriate prefix declaration).

   -- David
-- 
_________________________________________
David Durand              dgd@cs.bu.edu  \  david@dynamicDiagrams.com
http://cs-people.bu.edu//dgd/             \  Chief Technical Officer
     Graduate Student no more!              \  Dynamic Diagrams
--------------------------------------------\  http://www.dynamicDiagrams.com/
                                              \__________________________

Received on Thursday, 22 June 2000 11:55:40 UTC