Re: uri-comp draft necessary?

Jeremy Dunck wrote:
>> From: Paul Prescod <>
>> Miles Sabin wrote:
>> > ...
>>> Right, but this document uses five distinct namespaces,
> <snip>
>> Whether the TAG says it or not, choosing names that are non-equivalent 
>> by network infrastructure rules is merely common sense and I've never 
>> heard of a violation of that rule.
> I'm not sure of the context of this argument... It seems to be more 
> about namespace URIs than URLs as used on the web in general.  At any 
> rate, I'd caution against the assertion that character case doesn't 
> matter in URLs.  It happens not to in IIS, because IIS directly maps URL 
> heirarchies to its file system, and that underlying file system is not 
> case sensitive.  However, in Apache, and per the RFC (IIRC) URLs -are- 
> case sensitive, and therefore should not be 
> considered the same resource as, regardless 
> of what current improper implementations of web servers exist.

Ah, but DNS names are always case insensitive, so while the local 
path-portion of a URL MAY be case insensitive (from the server 
perspective), the DNS name (and scheme, if i am not mistaken) will 
always be case insensitive.  IDN may complicate this further, but i have 
not looked closely enough to be certain of this.

> Hopefully, you're arguing a different line, and I can stop worrying.  ;)
>> As far as "/" vs. "/default.asp", I'm 99% sure that IIS lets me take 
>> more fine grained control of that default equivalence if I need to do 
>> so. If I've been silly enough to deploy those two namespace URIs, I am 
>> probably not will just take a little bit more effort to 
>> disambiguate them.
> It does.  Any given file (in the current directory) can be made the 
> "default" for a particular directory, so that "/" and "/foo" can be the 
> same.
> As a side note, I'm pretty sure that a request to 
> "" is not even treated as equivalent to 
> "" by IIS... I think the former request is 
> redirected to the latter.

Yes, this is the case.  HTTP implimentations send 302 responses to 
useragents neglecting the trailing "/", where a directory match was 
found.  This provides a means of normalization by the server.  Not a bad 
approach, given that (presumablly) the server operator understands URL 
equivalance... perhaps the server should be tasked with any 
normalization, prior to any equivalence processing by the useragent?

>>  Paul Prescod
>  -Jeremy Dunck

--- peterd

<Quote type="random">
You can't have everything. Where would you put it?
<Author>Stephen Wright</Author>

PGP Fingerprint:
8994 8774 B682 3A04 B304  C4A2 D9DD 7E5B 8AAC 2D00

Received on Thursday, 19 December 2002 09:29:14 UTC