xml-dev - RE: simple question on namespaces. (more arguing)

[Mike Brown <mbrown@corp.webb.net>]

G. Ken Holman wrote:
> Why invent a new way in the Internet world for ensuring 
> unique values in as many possible domains in a coherent
> scheme?  There is already one to use

Well there are many other examples of syntax rules for values that are
supposed to be unique, and each ruleset is tailored for generating IDs
suited for a certain purpose. For example, there are competing syntax rules
for:
   - the values of ID-type attributes in XML documents
   - Message-ID values in email headers
   - language 'tags' like fr-CA and x-klingon
   - the name or location of a resource -- a URI

I do not see a good reason that URI syntax had to be the basis for
identifying namespaces. It should not be a surprise that confusion ensues
from this choice, since URIs were designed for a different purpose. They are
*not* generic identifiers; they are intended to be used to identify the
names and locations of resources. Whether the location or resource exists
and/or is accessible is irrelevant, as has been mentioned repeatedly. But
that doesn't change the fact that the URI identifies a location. Come on, is
something so abstract as a namespace really a *resource* that can have a
*location*? 

The URI spec says what the generic syntax is, and what URIs are generally
used for. The URN spec and various protocol specs refine the syntax and
intended usage a bit. I think the namespaces rec is the first one to come
along and say "OK, forget what you know about URI semantics; we don't care
about resource names and locations; we're just going to use the syntax for
this other purpose." Maybe it's not necessarily wrong to do this, but they
could have just as easily said they were going to use language tag syntax.
People would be just as confused, then.

I'm sure the post office has reserved a range of addresses for undeveloped
property in your city. What if I came along and said I'm mass-producing
widgets and need to give each one a serial number. I like the system your
post office used for producing street addresses, so I'm going to use that as
my serial numbering system. The IDs will read "7428 E. Main St.", "9683
Caldwell Ct.", etc. Oh, it's just a syntax, don't worry the fact that they
look like real street addresses for homes and businesses. When I put them on
my widgets, they're just for identifying the widget, that's all. What's all
the fuss about? Your post office doesn't guarantee that there will be
anything at every address it has on file, so why do you care if there's
something at the addresses  corresponding to my widget's serial number? I
just told you the serial numbers don't correspond to addresses at all. It's
just syntax!

You'd probably call me a few names, wouldn't you? :)

And actually, the syntax rules do not guarantee uniqueness of URIs, language
tags, Message-IDs, etc.  The namespaces rec actually says 2 separate things:
the identifier must follow a certain syntax, and 2 identical identifiers
refer to the same namespace. I would not call this a guarantee of uniqueness
nor a formula for generating unique IDs. I also don't see any guarantee in
the URI spec that a given resource can have only one URI.


   - Mike
____________________________________________________________________
Mike J. Brown, software engineer at            My XML/XSL resources:
webb.net in Denver, Colorado, USA              http://skew.org/xml/