W3C home > Mailing lists > Public > uri@w3.org > October 2003

RE: Names and addresses

From: Hammond, Tony (ELSLON) <T.Hammond@elsevier.com>
Date: Fri, 17 Oct 2003 11:20:34 +0100
Message-ID: <54A600C436EA694581B93E4BD4D4788A06B73CDE@elslonexc004.eslo.co.uk>
To: 'Larry Masinter' <LMM@acm.org>, 'Graham Klyne' <GK@ninebynine.org>, 'Gustaf Liljegren' <gustaf.liljegren@bredband.net>, uri@w3.org

> Let us imagine a policy where namespaces that have (interesting, useful)
> operational definitions get to be registered as URI schemes, namespaces
> that have a definition which establishes a (possibly delegated)
> naming authority / registration mechanism get to be registered as
> URN namespaces...

Allowing for the plausibility of the concept of URN (which becomes somewhat
problematic if the concept of URL is suddenly deprecated), then it must be
asserted that URN could be applied to multiple URI schemes and not just
pegged to a single URI scheme that so happens to call itself 'urn'. That
would be to introduce a point of gross non-uniformity into a system of
resource identification that champions its uniformity.

However, my personal view is that while I understand that the IANA would
like to keep everything nice and tidy and ship-shape and to distinguish
names from addresses, the reality is quite simply otherwise. The horse has
truly bolted (and been gone some time now I might add). It would appear to
be too late to retrofit these conceptual categorizations onto the URI
continuum. RDF and XML Namespaces, as two URI applications among others,
have fully put paid to that.

There is also a view that the 'cost of introduction of new URI schemes is
high'. I would take exception with this, and note that this view presumably
arises from the common expectation that URI schemes are necessarily
dereferencable and that transport protocols need to be supported. This is
not the case. The 'info' URI scheme, for one, expressly excludes dereference
for simplicity's sake. So what cost then to a URI processor? If it is aware
of certain URI schemes and the operations that can be performed upon those
URIs (e.g. retrieval), then fine. Otherwise it can profitably recognize a
URI for what it is in a given URI context and reliably use that for identity
purposes.

There is another aspect to this. Are the number of URN namespaces to be
likewise restricted? Or can they be allowed to multiply since evidently they
are apparently not 'interesting' or 'useful'. In short, are such namespaces
to be considered the lumpenproletariat of the URI world?

> ... and namespaces that have neither get to be debated endlessly
> on uri@w3.org, until the definers give up or else fit their schemes
> into one of the two camps.

By 'give up' I expect this is intended to refer to registration under the
IANA. The authors of the 'info' URI scheme already have existing
applications that will make good use of the 'info' URI scheme and have no
intention to 'give up'. (I assume BTW that the 'two camps' reference is to
'urn' URIs and not-'urn' URIs. Bit of an odd carve-up of the URI terrain.
But that's just a personal opinion.)

Tony


Tony Hammond
Advanced Technology Group, Elsevier
32 Jamestown Road, London NW1 7BY, UK

<tel:+44-20-7424-4445>
<mailto:t.hammond@elsevier.com>
Received on Friday, 17 October 2003 06:21:27 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 21:25:06 UTC