[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [idn] Registration rules and Canonicalisation rules
- To: Alan Barrett <apb@cequrux.com>
- Subject: Re: [idn] Registration rules and Canonicalisation rules
- From: James Seng <James@Seng.cc>
- Date: Wed, 23 Aug 2000 05:49:18 +0800
- CC: idn@ops.ietf.org
- Delivery-date: Tue, 22 Aug 2000 14:54:26 -0700
- Envelope-to: idn-data@psg.com
Excellent summary :-)
There is a thinline between what is considered character equivalent and
what is lexiconic equivalent. "color" and "colour" is lexiconic hence it
is not. Similar, han folding as I described in the PDF is also a
code-based equivalence.
I suggest we stick ourselves to only code or character equivalent and
leave lexiconic and context equivalent (which is beyond any reasonable
machine means) to administration.
-James Seng
Alan Barrett wrote:
>
> On Mon, 21 Aug 2000, RJ Atkinson wrote:
> > At 15:51 21/08/00, Keith Moore wrote:
> > >I don't see any way to make the registration rules "global" without
> > >imposing some language's assumptions on users of other languages
> > >which do not share those assumptions. and that IMHO is not acceptable.
> >
> > None the less, this WG had previously agreed earlier this year
> > that we needed to have the same canonicalisation/normalisation rules
> > apply all over in order to be interoperable.
> >
> > You would pose the question of whether we want to interoperate
> > globally or not. I believe the answer of that has to be that
> > global interoperability is mandatory, not negotiable.
>
> Ran and Keith don't seem to be communicating. Let me see if I can help.
>
> Keith says that "registration rules" probably cannot be global. Ran
> says that "canonicalisation/normalisation rules" must be global. I
> happen to agree with both of them about this, and there's no conflict
> here, because registration rules and canonicalisation rules are not the
> same thing.
>
> Registration rules are administrative procedures that prohibit one
> person from registering a name that is too similar to a name that
> somebody else has already registered. A silly example using the
> english/american language would be to say that the names "color" and
> "colour" may not be registered to different people. A less silly
> example using the french language would be to say that the names "cafe"
> and "caf<e+acute>" may not be registered to different people.
>
> Canonicalisation rules, in the present context, are rules about
> whether the DNS software thinks that two names (which might initially
> appear to be different) are equivalent.
>
> It's quite easy to imagine a world in which all DNS software thinks
> that "cafe" and "caf<e+acute>" have different canonical forms, yet
> some registries do and other registries do not permit "cafe" and
> "caf<e+acute>" to be registered to different people.
>
> --apb (Alan Barrett)