Re: uri-comp draft necessary?

Am Mittwoch, 18.12.02, um 19:17 Uhr (Europe/Berlin) schrieb Paul 
Prescod:

>
> Dare Obasanjo wrote:
>
>>  ... It would be difficult to
>> keep a straight face if the W3C TAG issued a document saying that
>>
>> http://www.example.com and HTTP://www.example.com
>>
>> were not equivalent then watching how that reacted with the notion 
>> that
>> namespace URIs should be dereferencable[0].
>>
>> [0] http://www.w3.org/2001/tag/ilist#namespaceDocument-8
>
> How would they react? So two different URIs happen to serve the same 
> document. Big deal. It happens all of the time on the 
> Web-as-we-know-it.
>
> These URIs all deliver the same data:
>
> http://www.microsoft.com/presspass
> http://www.microsoft.com/presspass/
> http://www.MICROSOFT.com/presspass/
> http://www.microsoft.com/PRESSPASS/
> http://www.microsoft.com/presspass/default.asp
>
> Nevertheless, I would be _extremely_ annoyed if my HTTP cache treated 
> them as the same URI just because at some level unrelated to URI 
> comparison somebody decided that it would be useful if they delivered 
> equivalent representations.

I think Dare's point was well made:

For HTTP servers and proxies, http://example.com/ and 
HTTP://example.com/ must
be equivalent URIs. They have to follow RFC 2396 in that. Additionally, 
the server in
99,999% of all cases will not have a choice as it will not see the 
scheme name in the
request.

As to your examples: I think any reasonable HTTP cache implementation 
would fold the second
and third URI in your example into one. So the cache would keep 4 
instances instead of 5. Otherwise
it would be interesting to add another chapter into the uri-comp draft 
for equivalence in HTTP
requests. ;)

>
> If allowing multiple definitions of "equivalence" is a fatal flaw to 
> an information system then the Web is already dead.

Yep. That's why: "we only need strcmp() for equivalence, so we can 
discard uri-comp draft" is
no good.

//Stefan

Received on Thursday, 19 December 2002 08:22:28 UTC