Re: TB16 Re: Comments on arch doc draft

On Wed, Jul 03, 2002 at 05:03:02PM +0300, Patrick Stickler wrote:
> Thus, using a mailto: URL to denote a person, or using a web
> page URL to denote a company, or using a namespace
> URI to denote both the set of terms and some document
> describing things related to that set of terms are all errors.

May I respectfully suggest that folks address this point of Patrick's?
IMO, this is an *extremely* common misconception, and it appears to
underly his entire argument.

Patrick believes that a URI should only identify one thing.  This is
absolutely true, and I don't think anybody here would disagree.  He
also believes that the bytes returned in response to a GET are
necessarily identified by that same URI.  This is not true.

For most GETs, it is the case that there are *two* resources being used;
the one on the HTTP request line (GET <some-uri>), and the one whose
content is returned as a bytestream representing the state of the other.
HTTP 1.1 provides the Content-Location header to distinguish between the
two; its value is the URI of the latter resource.

If Content-Location is not present, then there is an ambiguity.  But as
with all identifiers, what they identify is determined by use, not by
edict.  And invariably (I've yet to see a counter case), use decides to
identify the abstract thing, not the content.  To demonstrate this, pick
any URI, do a "backwards links" search from Google on it, and then take
a look to see the context in which its used.

(but note that Google has some nasty heuristics.  For example, it assumes
that http://www.yahoo.com/ and http://www.yahoo.com/index.html identify
the same resource).

MB
-- 
Mark Baker, CTO, Idokorro Mobile (formerly Planetfred)
Ottawa, Ontario, CANADA.               distobj@acm.org
http://www.markbaker.ca        http://www.idokorro.com

Received on Wednesday, 3 July 2002 14:35:46 UTC