Re: [idn] Fast nameprep vs. slow nameprep

[Patrik Fältström <paf@cisco.com>]

At 04.12 +0000 01-01-28, D. J. Bernstein wrote:
>Patrik writes:
>>  We need something between the keyboard and the config file which do the
>>  nameprep, so never non-nameprepped things end up in configuration files.
>
>I'm glad you agree. So why do you keep claiming that (for example) BIND
>and djbdns and Sendmail and qmail need to learn about nameprep?

I don't.

I have been talking about ACE v.s. UTF-8 (which is completely 
independent on where nameprep is done), and claim that IF we do 
nameprep and ACE in whatever tools that exists between the user and 
the software which handle the application protocols, then no changes 
are needed in the software you mention.

See the idna spec which myself and Paul has written.

The main idea in that spec is that you do nameprep and encoding as 
"high" up in the protocol stack as possible, as the nameprepped 
encoded domainname can be easilly reused in all protocols without any 
changes what so ever in qmail, sendmail, bind, djbdns etc.

It's even the case that the existing webpages which some people have 
created (or separate software) which converts between whatever 
charset one use natively and encoded nameprep is used, so the user 
know the ace encoded nameprepped string, then that string can be used 
in all existing software. Ugly, but doable.

I said one day on this list:

At 09.24 +0100 01-01-06, Patrik Fältström wrote:
>I.e. I could register one day maybe:
>
>   fältström.com
>
>...but could also on my buissness card have (if the ACE ends up being RACE):
>
>   bq--abtoi3duon2hf5tn .com
>
>Yes, it looks ugly, it explicitly is leakage which we all hate, but, 
>it  makes it possible for people with non-IDN software to access the 
>domain. Of course, I can in parallell have a domain which is 
>"faltstrom.com" or whatever, but I might think that my customers if 
>I help them that way never update their software. I.e. I _WANT_ 
>people to be able to use "fältström.com" probably, so I would only 
>help them to some degree, and that might be to give them the ace 
>encoding in paralel with the real name.

    paf