Follow-up to Monday's discussion of digits

[John C Klensin]

--On Tuesday, 18 November, 2008 07:53 -0600 Andrew Sullivan
<ajs at shinkuro.com> wrote:

> On Tue, Nov 18, 2008 at 06:45:47AM -0600, Eric
> Brunner-Williams wrote:
> 
>> Outside of the protocol, registries are free to implement a 
>> registry-local policy, which may restrict code points in a
>> label to one  range only, or one of two ranges, where one is
>> in the U+0030..U+0039  range, but not both of the ranges
>> U+0660..U+0669 and U+06F0..U+06F9.
> 
> Something that wasn't clear to me from the discussion
> yesterday is why this ought not just to be a registry policy
> issue, in the way we have said other possibly-confusing things
> ought to be.  Is the argument from potential for harm?

The key problem, as discussed during the meeting is to _enable_
registry policy by preventing combinatorial explosion.   In
other words, for most of the plausible cases in which there
might be IME problems and/or more than one set of digits are
used, the probably-best registry policy is to register all of
the numeral-containing digit forms used with Arabic script.

With this rule, that will typically require registering only two
labels for each proposed label that contains digits (European
digits and Arabic-Indic Digits  or European Digits and Extended
Arabic-Indic Digits) or three labels (European Digits and both
forms of Arabic-Indic Digits).  Without such a rule, one gets a
combinatorial explosion because one has to register all of the
possible combinations of digit forms for labels containing more
than one numeral.

Note that explanation is intended to clarify today's meeting
discussion for the mailing list.    Until Eric and I re-consult
our various authorities and then each other and report back to
the group, there is uncertainty about what the rule should be,
but the principle of avoiding a combinatorial explosion is
likely to remain.

     john