From: Kenneth Whistler (kenw@sybase.com)
Date: Mon Jul 28 2003 - 17:38:43 EDT
Michael Everson asked:
> >Because changing the canonical ordering classes (in ways not
> >allowed by the stability policies) breaks the normalization
> >*algorithm* and the expected test results it is tested against.
>
> Do you really think that algorithm with all its warts is going to be
> used 50 years from now? I really would like to know.
20 years, sure. 50 years is a little harder to prognosticate.
But then people have been uppercasing ASCII for 36 years now
with the following algorithm:
c = c - 0x20;
so just because an algorithm is flawed doesn't mean it won't be
used for a long, long time.
But in the long run, the "does it really matter in the long run?"
kind of argument isn't very convincing. You can always answer
that question in the negative, eventually, simply by adding
years to the span in question.
But Unicode normalization stability matters *now* for a large
number of implementations with significant investments. And
as for most decisions made about Unicode (and 10646), it is
what matters *now* and in the foreseeable short term which
counts the most for the decision makers. And it isn't that nobody
has longterm vision here, but when one of your goals is
longterm stability, you have to keep making shortterm decisions
which individually preserve that stability.
--Ken
This archive was generated by hypermail 2.1.5 : Mon Jul 28 2003 - 18:10:17 EDT