Demanding that every text-handling program deal with normalization is
analogous to demanding that every text-handling program deal with 8859-1
and KOI-R and so on. In contrast, in a better-engineered system, those
details are isolated inside a small number of programs.
Paul Smith writes:
Can you *guarantee* that non-normalised text won't ever get into emails?
That's analogous to asking ``Can you _guarantee_ that the text won't be
in 8859-1 or KOI-R?'' You're completely missing the point.
The question is about the costs of requiring normalized UTF-8
everywhere, as opposed to allowing unnormalized UTF-8. The extra
constraint has an obvious benefit for typical text-handling programs.
Keith is claiming, without a shred of justification, that the cost for
text-input programs outweighs the benefit for the much larger set of
text-handling programs. So far we haven't seen even one example where
producing normalized UTF-8 would be difficult.
---D. J. Bernstein, Associate Professor, Department of Mathematics,
Statistics, and Computer Science, University of Illinois at Chicago