W3C home > Mailing lists > Public > www-international@w3.org > January to March 2005


From: Martin Duerst <duerst@w3.org>
Date: Mon, 21 Feb 2005 09:24:47 +0900
Message-Id: <>
To: "Jony Rosenne" <rosennej@qsm.co.il>, <www-international@w3.org>

At 01:59 05/02/20, Jony Rosenne wrote:

 >> From: www-international-request@w3.org
 >> [mailto:www-international-request@w3.org] On Behalf Of Martin Duerst

 >> I think we all agree that from an user point of view, it would
 >> have been desirable to allow numbers at the end of RTL labels.
 >> However, the restrictions we were working with were the following:

 >> - Express the restrictions in terms of single labels, not
 >>    dependent on the rest of the domain name.
 >It seems the difficulty stems from this rule

Yes. We didn't set up this rule, it was the DNS experts which
very strongly insisted on it.

 >Probably the only solutions are either to relax the second rule, or to
 >accept ambiguities as per Mati's example and advise those registrars that
 >wish to support bidi IDNs that they ought to control some particular cases.

Unfortunately, the only solution currently is to accept the restrictions
as they are. There are many restrictions in domain names that do not exist
in naming in practice. The most obvious example is that names in the
wild contain spaces, but domain names can't contain spaces.

There may be a revision of IDNA in the future, but I don't know when,
and I'm not sure how easy it would be to bring in such changes.

 >Is the restriction only on Arabic (commonly known as European) digits, or
 >does it include also Hindi (commonly known as Arabic) digits?

The restriction is not on digits, but requires that a label containing
RTL characters both starts and ends with an RTL character. RTL
characters are characters of class R and AL. This does not include
Hindi/Arabic digits.

Regards,    Martin. 
Received on Monday, 21 February 2005 01:13:37 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 22:40:50 UTC