W3C home > Mailing lists > Public > uri@w3.org > February 2005

RE: New URI registration draft; significant changed

From: Dave McAlpin <Dave.McAlpin@epok.net>
Date: Wed, 23 Feb 2005 16:43:06 -0500
Message-ID: <1636C74D3A020E4781DC433A4D901D95798938@empire.dc.epokinc.com>
To: "Larry Masinter" <LMM@acm.org>
Cc: <uri@w3.org>

Good example. So if a scheme's ABNF doesn't re-use a lot of productions
from 3986, creating a separate ABNF for the IRI form isn't very helpful.
For a scheme like http it might be more useful, but 3987 already makes
it clear that to check syntax you have to first map to a URI, so maybe
an IRI version of the ABNF is never a good idea.

3987, though, gives the scheme designer certain options like

- Treatment of IDN components in normalization and comparison.
Application of RFC3490 and RFC3491, for example, or for which components
conversion to punycode doesn't affect equivalence. (Sections 3.1, 3.2.1
and 5.3.3 of IRI)
- Components to which BIDI restrictions apply (Section 4.2 of IRI)
- Whether BIDI restrictions in section 4.2 are strictly enforced for
various components of the IRI
- Case sensitivity for non-ASCII characters (5.3.2.1)

Most of these don't relate directly to the URI scheme definition, but
they have to appear somewher. What about an "IRI Considerations" section
as a standard part of scheme definition?

Dave

-----Original Message-----
From: Larry Masinter [mailto:LMM@acm.org] 
Sent: Tuesday, February 22, 2005 5:31 PM
To: Dave McAlpin
Cc: uri@w3.org
Subject: RE: New URI registration draft; significant changed

> I think I expected something stronger, like advice that the scheme's
> syntax should be defined in terms of both 3986 and 3987. I imagine
> future schemes including an appendix, possibly non-normative, with
ABNF
> based on 3987 for applications that natively support IRIs. I just
> wondered if it was too early to offer that as a best practice.


Well, let's take a current example:

http://www.ietf.org/internet-drafts/draft-duerst-mailto-bis-00.txt

is an example of a (revision of a) URI scheme definition that
is defined in terms of RFC 3986/STD 66, but consistent with
the advice in RFC 3987 (as recommended by draft-hansen-...).

Would you really want to see the BNF duplicated, once for URI
and IRI? I don't think it would be useful, and I think that it
would introduce opportunities for confusion or even conflicting
definitions.

Larry
-- 
http://larry.masinter.net


-- 

Checked by AVG Anti-Virus.
Version: 7.0.305 / Virus Database: 266.4.0 - Release Date: 2/22/2005
 
Received on Wednesday, 23 February 2005 21:43:13 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 13 January 2011 12:15:35 GMT