Re: Inclusions and other gotchas (was:Re: inclusion)

> Even that does not suffice.  There is simply no way based on URIs(*)
alone
>to tell if two URIs refer to the same thing or not.

I think that's overstated a bit. URIs define a way of recognizing when they
refer to the same thing.  It's just that this test isn't negatable.

Two URIs definitely occupy the same point in URI-space when they are
character-for-character identical.   Note that this is an abstract
statement about an address, and doesn't have anything to do with whether
there is anything at that address or not -- or whether the address returns
the same data twice in succession. (Consider a "hit counter", for example.)

Two URI References (remember the distinction)  definitely refer to the same
point when absolutizing yields two URIs which are the same
character-for-character.

HOWEVER: Failing those comparisons apparently does not mean that the two
URIs do _not_ refer to the same point in URI-space. For example,
http://www.ibm.com and http://204.146.81.99/ refer to the same point on the
web (until the DNS is changed, anyway) even though you can't prove that
from their URIs alone. The same problem arises if two different names are
aliased to the same address. It occurs again with character-escapes, which
the spec says are not considered when comparing URIs.

In other words: as far as I can tell from a much-too-brief scan, the
official test for "does URI1 equal URI2" only returns "yes" and "maybe" --
there seems to be no official way to answer "definitely not" without
attempting to dererence them, and even that doesn't seem to be completely
reliable.

Am I misunderstanding URIs? Is there in fact a clear statement somewhere
that "if we don't know they're equal, they may/must be treated as unequal"?


[I think I'm finally starting to understand what Tim is driving at, in
terms of the namespace declaration only declaring a _reference_ to a
(family of) point(s) in URI-space, rather than stating the actual "name" of
a specific Namespace. I'm still trying to convince myself that it's
actually practical (as opposed to possible or even desirable) to define
Namespaces that way. No answers yet, but some new questions...]

______________________________________
Joe Kesselman  / IBM Research

Received on Thursday, 25 May 2000 18:52:22 UTC