RE: [idn] Requirements: sorting

[Jonathan Rosenne <rosenne@qsm.co.il>]

I'll give an example: Unicode and 10646 both allow two ways to write an a
grave: a single combined letter (U00E0) or the sequence a (U0061) followed
by the grave accent (U0300).

The poor user, if his system support this character, has no control over
which form will be used. He just reads or hears about an interesting URL and
types it as he is used to type his own language, French for example. He may
have a key engraved with the accented letter on his keyboard, or he may have
used a "soft" keyboard, or he may have to type them separately, or he may
even have used a spell checker to add missing accents, in any case his
system will decide how to encode it.

So the iDNS would have a problem, unless it were decided that a domain name
must be normalized to a particular form.

Jony

> -----Original Message-----
> From: Randy Bush [mailto:randy@psg.com]
> Sent: Thursday, February 03, 2000 7:57 PM
> To: Jonathan Rosenne
> Cc: idn@ops.ietf.org
> Subject: RE: [idn] Requirements: sorting
>
>
> > I think it would be simpler to require "canonicalisation" by the browser
> > or user agent.
>
> i am confused.  please define canonicalisation.
>
> randy