[UA-discuss] Issue needs discussion and closure

[Asmus Freytag]

On 3/11/2018 11:59 AM, Andrew Sullivan wrote:
> On Sun, Mar 11, 2018 at 06:43:34PM +0000, Mark Svancarek via UA-discuss wrote:
>> Here is my reasoning:
>>
>> We forbid script mixing in the root for well-understood reasons.
> But we don't forbid script mixing in the root.  The LGR effort is
> busily trying to set the correct rules for this, but the root already
> has more than one script, and there are defintiely potential labels
> corresponding to Japanese words that require script mixing.  The rules
> are actually aimed at prohibiting mixing of _writing systems_, which
> in the last go round was approximated as "script".

There are some writing systems where use of the occasional ASCII 
character is
common; this is not supported in the Root.

There are some writing systems where mixing ASCII (or Latin) characters 
would
be of more concern than in others. Generally, when the script looks much 
like Latin,
as is the case for Cyrillic, such ASCII mixtures are problematic, while 
for Japanese
or Hindi they might not be.

>
>> The same thought process could be applied to second level and below.
> But this gets harder the lower in the tree you go, because there is no
> authority to enforce it.

There's the question of authority and there's the question of what makes 
sense,
or constitutes "best practice" in some general sense.

For the latter the intended audience for a zone would matter. If it is 
general, public,
world-wide or at least comprises users of more than one writing system, 
some of
the things developed for the Root would make sense whether as "starting 
point" or as
"recommendation".
>   Moreover, things that would be a very bad
> idea for the root, such as (say) Egyptian hieroglyphs, would be just
> fine at other layers of the DNS.  And there's the problem of different
> scripts in different labels, which is a permanent and unresolvable
> problem because of the nature of the DNS.
You mean having different labels in the same FQDN.
>> It's not an obligation, but the benefits to the user are the same and I think it is safe to say that it's a good practice to apply those same restrictions and exemptions to any label in a domain name.
> What is certainly safe to say is that you should not create
> identifiers where you don't understand what the implications are.
Nobody simply understands the implications. The Root Zone LGRs that have 
come
online and the ones still in the pipeline do embody such "understanding" 
for the set
of code points required for modern use in widely used scripts.

As documents they are reasonably detailed in documenting the issues and 
resulting
design decisions. They also  link back to general references so that it 
is possible
for readers to follow up those details, if inclined.

Most (by number) modern-use scripts need restrictions beyond defining a 
simple
repertoire. Developing these restrictions from scratch is a moderately 
complex under-
taking. Understanding and copying the ones developed for the Root Zone 
LGr is
much easier.

Not doing anything, is easier yet, but leads to very insecure IDNs...

A./
>   
>
> Best regards,
>
> A