Re: [idn] Thoughts on nameprep

["Bruce Thomson" <bthomson@fm-net.ne.jp>]

> But if two implementations do it differently, they will look up different
> domain names.
> That's the whole reason.

I understand that is the motivation for nameprep to repair "bad" names
into "good" ones. But my point was that at least for Japanese, and
I suspect for other languages as well, that anybody who has gotten
to the point where they can reliably enter a user name and password
in the language will already be able to enter a domain name.

Microsoft has recently been in the business of going beyond the spec
to rewrite domain names until they resolve. Adding left-off www.s
etc. A bit unsettling for purist geeks like me, but useful I suppose.
They also change lower case letters to upper case when I am typing
documents in MS Word, and I abhor that, because when it isn't
what I want, I can't turn it off.

Basically, it seems to me that correcting a user's mistake when he
types an illegal character in a domain is on the same level. A fuzzy,
friendly, nudge on the elbow to get the user a little further along.
But this is basically AI at work. Lots of heuristics, not too
predictable. It certainly isn't the kind of thing where people demand
consistency. If it works differently on a different browser or not at
all, so what?

This is especially important when you notice that people think that
nameprep is so complicated that it might not be possible to implement
it accurately in different resolvers. So there was even a suggestion
to move it into the DNS for for consistency. This is would be reasonable
if nameprep consistency was a high priority, but it isn't. Burdening
the DNS with this is asking for big trouble.

The important thing that needs to be in the idn spec is just the legal
character set. Even this is likely to change after it is in place. There
are just too many countries involved to expect to get it right the
first time. Priority should be on getting this nailed down, at least
temporarily, and getting this show on the road. Nameprep can
wait.

It would be wise, also, to learn from the mistakes made when defining
other recent specifications. Take, for example, the PCMCIA spec
(now called the PC Card spec.) The spec writers came up with
an extremely impressive, flexible spec that allowed interoperability
between a myriad of different cards and card readers. Unfortunately,
though, it was impossible to implement it completely on any given
product. So manufacturers would implement maybe 20% of it
and declare themselves compatible. The result is the computer
I bought yesterday, years after the spec was created, can only
handle a subset of existing cards. Thanks a lot.

Lets create just the minimum, and get maximum benefit.

Bruce

----- Original Message -----
From: "Harald Alvestrand" <Harald@Alvestrand.no>
To: "Bruce Thomson" <bthomson@fm-net.ne.jp>; <idn@ops.ietf.org>
Sent: Friday, March 09, 2001 11:41 PM
Subject: Re: [idn] Thoughts on nameprep


> At 12:39 09/03/2001 +0900, Bruce Thomson wrote:
> >Nameprep can and probably will be implemented at the
> >application level. Microsoft already prepends www. and
> >appends .com if I leave them off. They can handle this
> >kind of thing. But what is the pressure to put it in the spec?
>
> as you say, all implementations can, and probably will, handle "this kind
> of thing".
> But if two implementations do it differently, they will look up different
> domain names.
> That's the whole reason.
>
> --
> Harald Tveit Alvestrand, alvestrand@cisco.com
> +47 41 44 29 94
> Personal email: Harald@Alvestrand.no
>
>