Re: [9fans] A simple question

["boyd, rounin" <boyd@insultant.net>]

> Yes, that's where composition came into the picture.  But that
> needs to be clever, with character scaling to make room for accents
> becoming more than a trivial nuisance.  Technically, it doesn't
> matter what a symbol stands for, as much as it needs to be _presented_
> in an unambiguous, clear fashion.

yes that is not a problem.

> Whether it's a pronunciation issue or a distinct character (is the
> final letter in "papa" and "papà" a pronunciation aid or a different
> character in the sense of differentiating words with different
> meanings?) is not important to its internal or external repreentation.

but it's a big problem when you come to write sort(1).

> But I do get your point that overlaps of alphabets for different
> languages does add complexity.  Maybe there is enough scope in
> UTF-8 or Unicode to allow many-to one internal to external mappings.

it really is a nasty problem.  python [makes a crucifix sign] lets you add
codecs and all sorts of horrible crap.

> The existence of a phonetic alphabet is a different issue, too vast
> to address here (without composition capabilities, specially).

phonetic alphabets are easy.  well japanese would be perfect if
they had stuck with the kana, but the kanji really messes
things up.  i think the wost character has 17 readings and that
can ruin your whole day.

just my rounin tattoo:

    http://www.insultant.net/images/rounin.jpg

has three readings and a 4th in chinese.

> Suffice to say even in Plan 9 there are fonts that do not have all
> the useful glyphs in them, so whereas UTF-8 is a great abstraction
> for internal purposes, there should be a more definite standard
> about externalising it.

well maybe.  i'd be happier if i could chose my keyboard layout
as part of the install.

much as i hate lunix, the redhat install was pretty impressive,
but that's the 'army or programmers' tar pit.