IDNA2003 complication (was: Re: Changing the xn-- prefix)

[Andrew Sullivan]

On Thu, Mar 27, 2008 at 11:38:10AM +0900, Martin Duerst wrote:

> Can you give some details on what you mean by IDNA2003 complication?

Roughly, "All the things that the proposed WG seems to think it is
fixing."  

> I very much hope that we can fix IDNA2003 where it needs fixing, but
> I don't see, based on the current proposals, that overall, for an
> application programmer or for a registrant, things would get simpler
> with IDNAbis.

I agree.  But a proposal that we do _both_ IDNA200x and IDNA2003 means
that you have two sets of complications.  And they're disjunct.  For
instance, on the one hand, as a developer or operator aiming at
IDNA2003 conformance, I have to find some way of keeping around
ancient unicode libaries; whereas as a developer or operator aiming at
IDNA200x conformance, I have to worry about Unicode updates to make
sure my application accommodates them.  As an operator under ICANN
with an IDNA2003 system, I have to work out rules with various
language authorities (and finding the right one is a minefield all its
own) about what code points are in or out; whereas as an operator
attempting to conform with IDNA200x, national-language-authority
confusion may get solved (I'm actually pretending to be sanguine about
that; I confess that in reality I'm not quite so hopeful), because
there's an algorithm that determines the Internationalized-LDH-analogue.

I argue that, given that we already have strong reason to believe that
IDNA2003 is going to be deprecated, it would be a very bad idea to
specify it as some sort of minimally compatible protocol.  If you do
that, people will implement to the minimum.  I think it was John
Klenisn said in the BOF that, if we cannot accept that there will be
some possibly irritating backwards incompatibilities, we might as well
just stop immediately.  I think the corollary is that we cannot
specify a backwards compatibility criterion, or all we'll ever get is
that backward compatibility, because the cost of keeping around
two somewhat incompatible protocols is too high.

Best regards,
A
-- 
Andrew Sullivan
ajs at commandprompt.com
+1 503 667 4564 x104
http://www.commandprompt.com/