BIDI rules

[Eric Brunner-Williams]

Erik van der Poel wrote:
> On Thu, Sep 4, 2008 at 8:26 AM, Harald Tveit Alvestrand
> <harald at alvestrand.no> wrote:
>   
>> Erik van der Poel skrev:
>>     
>>> At the IETF meeting, I started thinking that we should not worry about
>>> getting the bidi rules exactly right this time, since we still have
>>> the opportunity to refine the spec as we move from Proposed to Draft,
>>> and then to Standard.
>>>
>>> But now I'm wondering whether we might remove the bidi reference from
>>> the IDNAbis protocol, and submit the bidi draft as an Experimental
>>> RFC?
>>>       
>> This would only be logical if we replaced it with a statement that RTL
>> characters cannot be used in domain names, unless within the confines of a
>> (documented) experiment - if the main documents were to remove all
>> restrictions, we can't put restrictions back.
>>     
>
> How about not putting restrictions back into the main documents? We
> could leave bidi out of the protocol, and leave it to zone admins and
> applications.
>   

How is this different from Harald's "RTL characters cannot be used in 
domain names, unless ... [some restricted scope statement]"?

> Bidi and spoofing are both *display* issues. If we can leave spoofing
> out of the protocol, why can't we leave bidi out of it?
>   

My prior life trying to write the bottom half of tty drivers and 
calculating the display size of tab and the directionality of backspace 
in nested bidi strings probably confuses me, but I can't see how the two 
issues are so similar as to be swept under the same rug.