The Future of IDNA

[Erik van der Poel]

On Thu, Mar 19, 2009 at 4:01 PM, Andrew Sullivan <ajs at shinkuro.com> wrote:
> On Thu, Mar 19, 2009 at 03:30:25PM -0700, Erik van der Poel wrote:
>> You haven't shown that they're equally important.
>
> That places the burden of proof the wrong way.  You want to introduce
> a restriction to the protocol (called "mapping").

IDNA2003 is a Proposed Standard. IDNA2008 is a set of Internet Drafts.
I realize that this particular WG's charter does not include mapping.
If that means we'd have to re-charter, that'd be really, really
unfortunate, but I'd rather re-charter than get it wrong (again).

> I argue that
> someone could easily prefer a different restriction in some other
> zone, and therefore this is a policy decision that needs not to be in
> the protocol.  It's often true that it would be more convenient for
> someone to include that person's policy in the protocol; that
> protocol-determined policy then comes at the cost of someone else's
> preference.  I don't need to show they're equally important; I merely
> need to show that there's a legitimate alternative preference available.

See my discussion with Ken about modern and classical Greek.

>> But seriously, the DNS is often used for names and words, and we're
>
> But it _doesn't_ use names and words.  They're labels.  Often they're
> not words.  "ns01" is not, I assert, a word, but it's a very common label.

ns01 and www23 are not the most visible labels on the Internet. We are
trying to internationalize the labels that ordinary users see most
often.

>> trying to internationalize the DNS. Why would we pay as much attention
>> to folks that want to distinguish Final and Normal Sigma in some
>> mathematical formula (if that is even a desire)? The DNS is not for
>> mathematical formulae, and it is not for novels either.
>
> It's not for words, either.  It's for identifiers.  Which could, in
> some cases, be distinguishable on the basis of this distinction.

See above.

>> > think have been outlined at length.  If you think those arguments for
>> > the problems are wrong, that's a different matter).
>>
>> Why is that a different matter?
>
> Because you haven't established that those arguments are wrong.  I
> haven't seen a solid argument yet that the initial problems which
> inspired the IDNAbis work are not still compelling.

The IDNA2008 rationale draft is not very compelling. It is quite
vague. Whenever Mark and I ask for explicit examples of confusion due
to mapping, John says something like: If we mention actual examples,
we will end up making character-by-character decisions.

> My reading of
> Paul Hoffman's draft is that there are too many negatives involved in
> the trade, and therefore we shouldn't do it, and not that the original
> diagnosis was wrong.

Of course some of the diagnosis was understandable. Turkish users
might be baffled by software that doesn't map dotted and dotless
letter i the Turkish way. Hopefully, client implementers will do a
better job with keyboard UI at some point.

But removing ALL mappings just because of that and a few other little
things? No, we have to weigh that against the other. And I tend to
agree with Paul that we need to keep mapping.

>> What remains to be done is to create some kind of forum where
>> interested parties can lobby for addition or removal of mappings.
>> Whether that forum operates by "rough consensus" or voting is another
>> matter.
>
> Lobby whom?

Maybe "lobby" was the wrong word. Maybe an ISO WG? Or a joint ISO/IETF
group? Joint Unicode/IETF group?

Erik