[XML-DEV Mailing List Archive Home] [By Thread] [By Date] [Recent Entries] [Reply To This Message]

RE: URIs harmful

• To: <bill.dehora@p...>,<xml-dev@l...>
• Subject: RE: URIs harmful
• From: "Danny Ayers" <danny666@v...>
• Date: Thu, 25 Jul 2002 12:45:20 +0200
• Importance: Normal
• In-reply-to: <000701c233a8$d448ac50$1fc8c8c8@mitchum>

>What are people's opinions on whether the authority which created the
>URI can/should change which thing is being named?

Great question!

In an ideal world the mapping between URI and thing named, once created
should remain constant, otherwise the whole identification business is
pointless.
This leads into the familiar dread region of what is actually being
identified, which when talking of uri+url could be a document or what is
returned by applying a certain algorithm to the identifier or numerous other
interpretations. Permanent mapping of URIs is very close to the purl idea,
but this has drawbacks - e.g. the clunkiness of datestamping the path for
versioning (ugly & not easy to index by other means). Then there's the whole
'expires' business.

The key I suppose is in the semantic interpretation :  http://www.w3.org is
the W3's home page (and namespace), the content of the document there may
vary. The only way to square the circle is with metadata. For instance,
being pragmatic I think we have to assume all URIs have a default property
of 'transient'. If we're talking about something like a spec, then somewhere
there should be information available describing the temporal and webspacial
range of the resource mapped by the URI. This way the mapping remains
constant, even if the resource that is the object of the mapping turns into
a frog.

A minor problem here is that interpretation of URIs by current agents is
generally quite tightly coupled. I don't think this is a major issue in the
sense used in the spurious 'expectation' argument regarding http: URIs - if
the XML namespace is shown as a link in an editor or browser, that's the
agent's fault, not the spec's. Backwards compatibility with broken systems
isn't a good idea. However, for the transition to the Semantic Web to be a
smooth as possible, the specs need to be as unambiguous as possible, and
(IMHO) loosely coupled. It shouldn't be necessary to implement http: to use
URIs with this prefix (it shouldn't be necessary to implement XML Schema
just to publish (X)HTML pages, but that's another story...). A layering in
which upper layers assume a set of defaults (like 'transient') which reflect
the current state of the web I think is also desirable.

Cheers,
Danny.

• References:
  • RE: URIs harmful
    • From: Bill de hÓra <bill.dehora@p...>

• Prev by Date: RE: DNS based URIs that don't imply access method
• Next by Date: Re: stupid URI tricks was: Re: URIs harmful
• Previous by thread: RE: URIs harmful
• Next by thread: Re: URIs harmful
• Index(es):
  • Date
  • Thread