Re: Documents, Cars, Hills, and Valleys

On Mon, 29 Apr 2002, Aaron Swartz wrote:

>> If you say that an HTTP header identifies a car, and then a GET returns a
>> picture, how do you refer to the picture?

> We have to do this already! The thing you get back from an HTTP GET changes
> over time and based on content negotiation.

Ok, what if you don't want to talk about the bits, but the picture in the
abstract?  Just like you might want to talk about a web page or document
in the abstract. How do you distinguish the two?

It is at this point, though, that those of us on the side of "http URIs
shouldn't be used for non-network retrievable resources" have a problem.
The abstract documents or pictures aren't retrievable.  However,
intuitively I know that there is a big gulf between things that are
primarily composed of bits and those that are composed of atoms.

For example, it seems perfectly reasonable to me to use an http: URI to
denote a picture I took with my digital camera, and it is even better if
that URI points at a server which can serve it up.  Likewise, it seems
reasonable to retrieve this email over http, since it is now and always
has been composed of bits. However, it seems problematic to me to use such
a URI to denote the Mona Lisa, because no matter what I get back from an
http request, it sure ain't going to be the Mona Lisa (hopefully, it will
be a photograph of it).  I'm still trying to decide what category the
Declaration of Independence falls in.  Perhaps a continuum of
"retrievability" is in order.

Anyone care to help me formalize this intuition?

Nick Matsakis

Received on Monday, 29 April 2002 20:34:15 UTC