Namespaces are Simply Punctuation (was Re: [namespaceDocument-8] What should a "namespace document" look like?)

Hi folks,

Sorry to just jump into the middle of this...

I've been chewing on this "what is at the end of a namespace"
issue for some time, and the conclusions that I have come
to are as follows:

namespace != vocabulary

  One may have multiple vocabularies grounded in the same

  Thus, there can be a 1:N relation between namespace
  and vocabularies.

namespace != doctype

  One may have multiple doctypes using vocabulary terms
  grounded in the same namespace and associate different
  content models to the same term.

  Again, there can be a 1:N relation between namespace
  and doctypes.

  Thus basing the interpretation of an XML instance on
  the qualified name of the root element does not actually
  garuntee proper interpretation.

namespace != digital web resource

  A namespace URI need not resolve to anything. It can be
  any valid URI which means it can be a URI for an abstract
  concept. It's only relevant quality is that it is globally

  Thus, there are namespaces which will never, can never,
  resolve to e.g. a RDDL instance or any other data stream.

  There is no such thing as a "namespace document".

namespace == punctuation

  A namespace is nothing more than punctuation. It prevents
  local names from colliding in a context of global syndication
  of names.

  One cannot presume nor assert that a namespace denotes anything;
  not a vocabulary, not a doctype, not a web resource. Once used
  as a namespace, that URI ceases to have interpretation as a URI.

Now, that said, I think that RDDL is moving in the right
direction. A directory/catalog of knowledge is what we
need -- it's just hooked to the wrong star, so to speak.

I think what we really need is something akin to good old
SGML catalogs and public identifiers. Perhaps RDDL recast
in RDF and public identifier URNs which resolve to
catalogs, or such.

We will continue to have no end of trouble and confusion
if we continue to allow folks to think that namespace URIs
are *real* URIs. They're not. They're just punctuation.

Yes, we need a catalog like solution, but let's please find
a more accurate and proper star to hook our solution to, yes?

Let's leave namespaces to do their simple job as puctuation
and move on.



On 2002-02-05 22:34, "ext Tim Bray" <> wrote:

> At 02:30 PM 05/02/02 -0500, Ian B. Jacobs wrote:
>> Summary: The section on namespaces in Web Architecture
>> from 50,000 feet states: "The namespace document (with the
>> namespace URI) is a place for the language publisher to keep
>> definitive material about a namespace. Schema languages are
>> ideal for this."  Tim Bray disagrees.
> I only disagree with the last 6 words, about schema languages.
> The rest I agree with 100%, and am on the record about the kind of
> resource that might be useful as a namespace document: see
> By the way, this is closely related to our issue (rdfmsQnameUriMapping-6)
> -Tim

Patrick Stickler              Phone: +358 50 483 9453
Senior Research Scientist     Fax:   +358 7180 35409
Nokia Research Center         Email:

Received on Wednesday, 6 February 2002 02:23:10 UTC