[
Lists Home |
Date Index |
Thread Index
]
2/16/2002 9:40:13 PM, "Simon St.Laurent" <simonstl@simonstl.com> wrote:
>So resources are kind of like Hegel's Spirit?
I enjoyed Frank Willison's reaction to the XML DevCon conference
last year too ... but URIs as REST defines them seem like a
very powerful, not useless, abstraction. Let's think about a "resource"
for students that consists of common sense interpretations
of Hegel's concepts. That "resource" is definitely more
than a collection of "documents" authored by Hegel; it could be
some sort of Artificial Intelligence that has been programmed/trained
to answer questions about Hegel's philosophy,or maybe a
bunch of unemployed humanities Ph.D's sitting at a computer,
interprerting URIs and typing responses.
Let's say that by GET-ing the URI
http://www.hegelfordummies.com/ask?explain+"geist"
you will get a nice XML document explaining what the
Universal Spirit is all about, an error message saying that
the question is unanswerable, or -- if there are humans back
there -- an XML message suggesting you try again in a couple of
hours and it should have the answer).
> I find it hard to believe that anyone but architects wants to
> see architecture built on identification rather than something slightly
> more tangible - say, location, even an abstract location in a computer.
I would argue that hardly anyone would want to know anything but
the fact that by forming URIs that describe questions about Hegel
we get back a representation of an answer to the question.
We don't know if the "resource" is a static document, a call into a
program, or a message to a human. Thus, the URI is a very abstract
identification of a document that could exists, not a pointer
to a physical location in a filesystem or database.
But this is a *useful* abstraction; a Desperate Philosophy
Hacker doesn't care what it it refers to physically, he/she just
wants an answer, and GET-ing a URI in the immense space of possible
questions about Hegel produces an answer.
As a less contrived example, consider
http://www.amazon.com/exec/obidos/ASIN/0764547771/qid=1013913357/sr=1-5/ref=sr_1_5/002-8423033-3456844
Is that a physical location on Amazon.com's server? A database query? (probably!)
A message to a guy who takes your book off the shelf and scans the cover?
(on busy days, I sometimes wonder!). Do you care, as long as you get
the representation you asked for?
I don't want to get into namespaces and URIs; there be dragons,
And I appreciate that URIs are somewhat underspecified with
respect to comparison rules, consistency across schemes, etc.
But there is an awful lot of REST-ful baby in that URI bathwater, so
be VERY careful about suggesting that we would be better off
thinking more concretely. For example, I presume that Fielding did
not worry about "Second Generation Web Services" when considering
the potential of REST 8-10 years ago, but it is the *abstraction*
inherent in the definition of "resources" and "identifiers" that lets us
do so today.
|