Re: Are *relative* URIs as namespace nemes considered harmful?

At 02:45 AM 5/17/00 -0400, Tim Berners-Lee wrote:
>Using URIs gives namespaces incredible leverage.
>Firstly, for each URI scheme, the questions of identify properties which
>become ratholes here have been explored in great depth. You have a
>rich choice of schemes to chose from for identifying namespaces.
>Secondly, a namespace becomes something which an be connected
>to other designs.  We will want to link to them, comment on them,
>state assertions about them, maybe also intellectual property rights,
>compatability information, and so on and so on.  Using URIs
>allows namespaces to gain from all that work, allowing the power
>of hte web technology to grow in a multiplicative rather than
>additive way.   This is not just for the semantic web ideas
>but for the web in general.

If there was any kind of consensus on what URIs could or should point to, I
might find this argument convincing.  I've noted namespace URIs as a
possible tool for pointing to XML packaging, for instance (see
http://purl.oclc.org/NET/xpdl).

That consensus has never emerged - in the case of XHTML, it proved
extraordinarily divisive.  In its absence, I've strongly recommended that
users treat namespaces as opaque names, merely taking advantage of DNS and
URI structures to keep us out of each other's way.

In some ways, I suspect that it's far too late to use URIs in general
within namespaces as pointers to resources.  There are far too many
namespace URIs that point to nothing at all, and those that do point to
something point to a set of vividly diverse resource types - DTDs, XDR, XML
Schemas, HTML (which may or may not describe the namespace at all).

Making namespace URIs work effectively in this anarchic context is going to
take, I suspect, some modification or addition to URIs.  A new protocol
perhaps, called 'package' or 'schema' for instance, that gives an
application a much better idea of why it should bother retrieving
information from the location specified.

I suspect that a lot of folks might not like this interpretation, but I
don't think the current range of approaches is likely to coverge on
something reliable otherwise.  For this discussion, its implication is just
that such dreams shouldn't be tied to the relative URI question - absolute
URIs may be necessary for meaningful non-string-matching uses of namespaces
anyway.

Simon St.Laurent
XML Elements of Style / XML: A Primer, 2nd Ed.
Building XML Applications
Inside XML DTDs: Scientific and Technical
Cookies / Sharing Bandwidth
http://www.simonstl.com

Received on Wednesday, 17 May 2000 09:41:55 UTC