Changing DISALLOWED (was Re: Reserved general punctuation)

John C Klensin klensin at
Thu May 1 18:43:06 CEST 2008

--On Thursday, 01 May, 2008 09:20 -0700 Paul Hoffman
<phoffman at> wrote:

> At 12:04 PM -0400 5/1/08, Vint Cerf wrote:
>> Coming back to UNASSIGNED and DISALLOWED, is it correct to
>> say  that  all UNASSIGNED code points are DISALLOWED (since
>> allowing them makes  no sense if the code point has no
>> character associated with it).
> Not at all. See section 1 of Patrik's protocol document. The
> two are completly disjoint.


Assuming that we have a very crisp definition of "unassigned",
the difference is that a code point can move from UNASSIGNED to
any of the other categories just by assignment of a character
and its properties.  No special IETF action (committee, RFC, or
otherwise) is required.  For any of the other categories, some
specific IETF consideration and action is required (how
significant that should be is really a separate question,
although it has obviously gotten entwined with this one).  This
distinction is at the very core of the "version agnostic"

I think Patrik's proposed change to "tables" to test
"unassigned" and assign that category first solves the problem
with the crisp definition.   If we were going to make any
classification assumptions based on the properties of code
points to which no character had yet been assigned (see Mark's
notes and Paul's followup), we would need to be certain that all
of those properties were absolutely immutable (in the
plain-language, as well as Unicode, senses).   While that might
be possible for some properties (e.g., block membership) and
some code points, it does not appear to me to be generally
possible and hence leads us into all sorts of places where we
don't want to be.


More information about the Idna-update mailing list