[
Lists Home |
Date Index |
Thread Index
]
Joshua Allen wrote:
>
>...
>
> It depends on how you define "representation" and "might" :-)
> Everything *could* have a representation someday, so that's really not
> justification for using http: identifiers. IMO, if you feel that
> representation retrieval (via synchronous HTTP GET) is likely to be an
> important function of the resource, then it makes sense to use http:
> identifier.
The word "important" is very vague. What resources do you have work with
that you feel confident will never need at least one of prose
documentation or an RDF description? If you are not confident, shouldn't
you hedge your bets?
>...
> Most people wouldn't want to interact with beaches via http.
They want to interact with representations of beaches via HTTP.
Especially prose and RDF representations.
> (It's more likely that there would be several http: identifiable sites
> which talk *about* a particular beach, and it would be better for
> everyone involved if they did NOT use their http: URI as the identifier
> for the beach. Consider the following example:
>
> A) Site: http://www.tybeebeach.com says
> "urn:beaches:ga-tybee qualityIs great"
> B) Site: http://www.tybeega.org says
> "urn:beaches:ga-tybee qualityIs poor"
> C) Site: http://www.mypersonalsite.com says
> "http://www.tybeega.org qualityIs poor"
>
> This use of identifiers has two important characteristics:
>
> 1) The use of a neutral identifier allows the sites to share the same
> identifier and permits people to aggregate metadata from many places.
There are no neutral identifiers. Somebody has the authority to assign
within a namespace and someone does not. If Microsoft and Oracle cannot
agree on a neutral identifier then they can ask the W3C or Oracle to set
one up. Whether HTTP protocol or otherwise.
> For example, a crawler like google could crawl both sites in A and B,
> and allow people to search for reports on "qualityIs" of the particular
> beach, without regards to who owns the individual http: sites.
An HTTP URI would have this same property. It is, after all, just a
string in a slightly different syntax.
> 2) Keeping the "web site" different from the "beach" permits people to
> make assertions about the "web site" as in C.
An HTTP URI could have this same property. You can set up an infinite
number of them, even new ones representing beaches separate from "beach
web sites." You could just have the text of the representation say:
"This page represents such and such a beach."
So HTTP URIs have the characteristics you outline *plus* they are easy
to dereference for prose, RDF or whatever else is a good representation
for the resource.
--
Come discuss XML and REST web services at:
Open Source Conference: July 22-26, 2002, conferences.oreillynet.com
Extreme Markup: Aug 4-9, 2002, www.extrememarkup.com/extreme/
|
