Re: [idn] Openendedness of Unicode/10464 (was: RE: [APTLD iname 24] Re: [idn] Proposed suggestions from AsiaPacific Top LevelDomain meeting)

[James Seng <jseng@pobox.org.sg>]

Actually, after spending some time thinking, I think it might be better not to
restrict the characters allowed in the requirement document. Originally I
thought that this might be a simple issue but i was very wrong so I will eat
my word.

I think Kent's suggestion make sense, ie we should not disallowed any
characters or treat any specially. Instead, it is up to the operational to
decide what to allowed and whats not, perhaps an a new RFC to supercede
RFC1123.

-James Seng

Karlsson Kent - keka wrote:
> That's why I said that which characters are allowed or not is a registration
> issue.  There must be nothing in the IDN implementations themselves
> that 'police against' not yet allowed characters, because such 'policing'
> causes problems for names that SHOULD be allowed later on.  The
> *registrators* should police against registring names that include
> unallocated
> code points, *until* those code points are allocated in Unicode/10646.
> Non-identifier characters (like symbols and punctuation, except a small
> number of them) should be policed against *by the name registries*,
> but *not by IDN implementations themselves*.  They may be useful
> later, and should be 'reserved' for that.  They must not be 'illegal'
> in the sense that domain name servers reject them (in any other
> way than "not found" after a diligent search).
> 
> Yes, I know, this WG is not about what registrators are allowed to do,
> but this issue falls somewhere inbetween 'protocol' and 'policy'.
> 
>                 Kind regards
>                 /kent k