wrt IDNA2008 migration (was: IDN processing-related security considerations for draft-ietf-websec-strict-transport-sec)

[John C Klensin]

--On Saturday, October 01, 2011 17:27 -0700 Mark Davis ☕
<mark at macchiato.com> wrote:

> The third choice, of course, was to maintain backwards
> compatibility with 2003 for all characters in Unicode 3.2, and
> just extend the same principles to new characters. That is a
> much easier migration path...

Unfortunately, that level of compatibility would have
permanently doomed those who believe that ZWJ and ZWNJ (and some
other "map to nothing" cases) to permanently living with
distinctions that they consider very important.  Perhaps that
would have been the right answer but, again, it wasn't what the
WG decided on.  

> We've seen this before. XML 1.1 only had a small
> breaking-compatibility changes, but those changes were enough
> to completely doom it.

Although the jury is still out, the same could be said for IPv6.
To the extent that analogy is relevant, one might suggest that
there weren't enough perceived benefits in XML 1.1 and IPv6 to
justify dealing with the incompatibilities... or the conversion
and deployment costs whether or not the incompatibilities
existed.  It is harder to prove that the incompatibilities alone
are to blame.

On the other hand, if strict backward-compatibility with
deployed technologies were always the right answer, we wouldn't
need to be having discussions about conversions of
tungsten-filament incandescent lighting to CFLs and LEDs because
we would still be using candles and gas or, to use an older
analogy, we would be worried about a rather different sort of
pollution and waste problem in our cities than emissions from
automobiles.

    john