Re: [I18nrp] Last Call: <draft-faltstrom-unicode11-05.txt> (IDNA2008 and Unicode 11.0.0) to Informational RFC

["Asmus Freytag (c)" <[e-mail]>]

On 12/6/2018 8:43 PM, Patrik Fältström wrote:
> On 7 Dec 2018, at 1:59, Larry Masinter wrote:
>
>> Registries should be conservative in the sense of advising their clients about the likely usability of the registered name among their target audience.
>> I don't think the IDNA RFCs currently are conservative enough.
> Agree. Specifically, as so many are misunderstanding what the RFCs are saying, and even if also others (like SSAC) request conservatism when coming up with a subset of PVALID (etc) repertoire, I felt it can never be said too many times this is a requirement.
>
>> As I've noted before, usability includes "re-enterable" (including "retypable") .
> I might not agree with this, as for example very few characters in arabic script are "retypable" for me. So this is at least a separate issue. But what the registry chooses, I feel is up to them, as long as they do come up with a good repertoire.


And there you are already limiting yourself to a tiny corner of the problem.

The list of code points - in isolation - is the smallest issue.

For most scripts (by number), the question of context rules for the 
entire label (built from context rules for syllabic context of certain 
categories of code points) are what distinguishes gibberish (stuff that 
neither renderers nor human readers can process) from structurally sound 
labels representing words or the kind of non-word mnemonics applicable 
to that script.

For some scripts, a registry policy that does not establish variants, is 
either not conservative or does not match deep-seated user expectations. 
For zones that support multiple scripts, the experience of the Root Zone 
is showing, almost all scripts need robustly defined cross-script 
variants to make the zone minimally robust.

Defining context rules and variants is the meat of the problem of 
defining conservative, secures and usable registration policies.

All the talk about "repertoires" hides this fundamental truth - which is 
the more misleading as registries for some well-known European languages 
may survive without the other two pillars or LGR design - as long as 
they support a single language (or a few closely related ones) only. 
These languages being familiar to many in this discussion may reinforce 
the misunderstanding that repertoire selection is primary.

It is not - it just is the first step.

A./