Mixtures of scripts

[Kenneth Whistler]

Paul Hoffman wrote:

> At 9:45 AM -0800 12/18/06, Mark Davis wrote:
> >I don't think it is necessary. If mixtures of scripts are not 
> >displayed (eg the user-agent flags them as discussed before), then 
> >they are not a problem. If mixtures of scripts *are* allowed, then 
> >there are so many other problems (eg with Cyrillic) that these pale 
> >in comparison.
> 
> There is nothing in the current version of 
> draft-klensin-idnabis-issues that prohibits "mixtures of scripts".

I think Mark is fully aware of that. His statement needs to
be interpreted as "If display of mixtures of scripts is
allowed in user agents, then..." The point is that Latin/Cyrillic
is a *big* problem. Most of the other script mixing issues that 
people worry about pale by comparison to that. What *that* means
is that they don't contribute much to the weighting of the
cost/benefits of various solutions. If we can satisfy ourselves
that we have a system that works for Latin and Cyrillic, the
rest of the scripts pretty much just come along for the ride.

> Please note that this says "might", and that it is about registries 
> and not the protocol itself. Thus, assuming that "mixtures of 
> scripts" are not allowed is incorrect.

I agree, and I think Mark and Michel do, as well.

> 
> If someone wants to this protocol to disallow mixtures of scripts, we 
> need to see the specific way that will be done. Until then, we should 
> assume that the current IDNA rules about mixing (namely: none) are in 
> effect.

Correct. Michael Everson seems to be the one pushing to
disallow mixture of scripts at the *protocol* level. I don't
think that is either practical or necessary, and trying to
enforce it now through a revised StringPrep (or some other
means) would be more of a backwards compatibility problem
than the kinds of restrictions we are already trying to
place on the protocol to disallow unneeded characters.

--Ken