W3C home > Mailing lists > Public > public-html@w3.org > September 2007

Re: Spec review: "URI (or IRI)"

From: Julian Reschke <julian.reschke@gmx.de>
Date: Sun, 02 Sep 2007 20:46:15 +0200
Message-ID: <46DB04F7.8090407@gmx.de>
To: Robert Burns <rob@robburns.com>
CC: "public-html@w3.org" <public-html@w3.org>

Robert Burns wrote:
>> Can you elaborate where the distinction is meaningful in the context 
>> of HTML?
> Actually, I think you just did elaborate :-) The xmlns: attributes take 
> an IRI. Those IRIs include by definition a URLs. They also include URNs 

Actually, it takes a URI reference (see 
<http://www.w3.org/TR/2006/REC-xml-names-20060816/#concepts>). So this 
includes references (although discouraged), but not IRIs (well, not IRIs 
that do not happen to be URIs as well).

> and either 'names' or 'locators' expressed through Unicode characters. 
> However, its not clear what using a URN means for an href attribute. If 
> that URN is defined using a scheme that defines a transport protocol, 
> then a resource could indeed be located. Are there any examples of this. 
> Even in my proposal to use URNs for the @cite attribute, the idea is to 
> resolve those URNs (or IRNs) into URLs (or IRLs) before retrieving the 
> located resource. I suppose something like the "tel:" scheme is a bit 

The cite attribute is a good example for a case where having a unique, 
but not directly resolvable identifier (such as an ISBN URN) is better 
than not to have it. Of course, it's even better to have something that 
*can* be resolved. Whether this is the case however is not controlled by 
the document author, or the HTML spec, but by the user agent (and the 
operating system it runs on). So I really doubt that the distinction is 
always meaningful.

> more ambiguous: is that a name or a location appended to that scheme? 

I don't see how "tel" would be different from "mailto".

> However, there are clearly IRIs that are locators and when an attribute 
> takes IRIs that are not locators I think we should say something about 
> what that means (as we can easily with xmlns: attributes or as done in 
> the earlier mentioned proposal[1]).

Yes, if it's only role is to *name*, then it doesn't hurt for the spec 
to state that.

Best regards, Julian
Received on Sunday, 2 September 2007 18:46:40 UTC

This archive was generated by hypermail 2.4.0 : Saturday, 9 October 2021 18:44:21 UTC