[9fans] combining characters

[9fans] combining characters

[rog]

erik wrote:
> rob has suggested passing uncomposed characters to libdraw and handling
> the problem there.  but there's one problem with that.  how do you stick
> a nonspacing horn onto an arbitrary letter?  how do you put a grave accent
> on top of that?  (transliterations of cryllic to the roman alphabet use some 
> double- and triple- accented letters which do not exist in precombined form
> within unicode.)

perhaps there are actually two problems here:
1) how to get libdraw to map back from a sequence of combining characters
to a character in the font that represents that sequence.
2) how to draw sequences of combining characters that don't exist in precombined
form within unicode. it's quite possible that one might wish to provide
pre-rendered glyphs for some of these sequences - the current font format
can't deal with that.

another issue is dealing with code (e.g. libframe) that assumes that
characters do not overstrike - i.e. that there's a 1-1 correspondence
between Runes and glyphs.

yet another is how one should deal with character-based indexing, for instance
indexing in sam expressions - does /é/-#0+#1 point to the character after
the unadorned e, or after the whole sequence?

it'd be nice to sort this issue out properly; surely it shouldn't be
too hard?

Re: [9fans] combining characters

[quanstro]

On Fri May 19 09:38:23 CDT 2006, rog@vitanuova.com wrote:
> perhaps there are actually two problems here:
> 1) how to get libdraw to map back from a sequence of combining characters
> to a character in the font that represents that sequence.

this is pretty easy.  the unicode standards provides cannonical compositions.
i think it would be easier for libdraw to insist that string be given strings that
have been cannonicaly composed.  perhaps a job for tcs.

> 2) how to draw sequences of combining characters that don't exist in precombined
> form within unicode. it's quite possible that one might wish to provide
> pre-rendered glyphs for some of these sequences - the current font format
> can't deal with that.

the general case doesn't seem like it would yield a solution with a bitmap font.
sure you could put a circumflex on an "a".  but what about dashed letters like
ł?  drawing a dash through an arbitrary character gets to be a real pain.  

the good news is that solving #1 would take care of most problems.  unfortunately,
some romanized versions of russian and vietnamise (i believe) would still not work.
but we would get 80% of what we would like without the pain of trying to treat
a bitmap as if they were vector character descriptions a la metafont.

> 
> another issue is dealing with code (e.g. libframe) that assumes that
> characters do not overstrike - i.e. that there's a 1-1 correspondence
> between Runes and glyphs.

charofpt would be a problem.  there would be some problems with picking a proper
endpoint for highlighting.  a break between the base and the combiners would
be a problem.  i think the largest problem here would be dealing with the character
height.  currently in libdraw a character's height is the font's height. this isn't true
for many fonts we already have -- ÄÖÜ☺ tend to get clipped with pelm because they are 
taller than the font file claims. just expanding the height of the font would look pretty 
funny in the absence of taller characters.

> yet another is how one should deal with character-based indexing, for instance
> indexing in sam expressions - does /é/-#0+#1 point to the character after
> the unadorned e, or after the whole sequence?

thair be dragons here.  the library of congress has a 100-page manual on alphebetization
of languages with roman letters.  different languages have different rules (sometimes for the 
same codepoint); a language sometimes has different rules for different codepoints.
then there are ligatures.  in german ss and ß are sorted the same. 

there are probablly only two sensible ways to deal with this.  either strip/do not strip
all combiners and do a naive sort or define some sort of locale.

> it'd be nice to sort this issue out properly; surely it shouldn't be
> too hard?

i believe this is another entry for the "famous lies list," ranking somewhat
below "check's in the mail" and above "i have this friend who...."

- erik

Re: [9fans] combining characters

[quanstro]

>> yet another is how one should deal with character-based indexing, for instance
>> indexing in sam expressions - does /é/-#0+#1 point to the character after
>> the unadorned e, or after the whole sequence?

>thair be dragons here.  the library of congress has a 100-page manual on alphebetization
>of languages with roman letters.  different languages have different rules (sometimes for the 
>same codepoint); a language sometimes has different rules for different codepoints.
>then there are ligatures.  in german ss and ß are sorted the same. 

uff.  this answer doesn't fit the question.  i think base+combiner* should be treated as 
an indivisible character.  but again, if we use cannonical compositions, this case can be
avoided except in cases where the character can't be drawn anyway.

- erik

Re: [9fans] combining characters

[Jack Johnson]

On 5/19/06, quanstro@quanstro.net <quanstro@quanstro.net> wrote:
> On Fri May 19 09:38:23 CDT 2006, rog@vitanuova.com wrote:
> > it'd be nice to sort this issue out properly; surely it shouldn't be
> > too hard?
>
> i believe this is another entry for the "famous lies list," ranking somewhat
> below "check's in the mail" and above "i have this friend who...."

My boss (actually my boss' boss) likes to chase this idea: is this
special cause or common cause?  Are we fixing a broken system, or are
we breaking a working system to try to solve a problem for only 3% of
the cases?

It sounds like there's definitely room for improvement, but I think
it's a rabbit hole (with Alice in tow) to find the one general
solution out there for the entire problem space.  Would it be great to
find that solution?  Most definitely.  But that could take a lifetime
or more of work.

So for me, the big question is what's the most graceful way to get
from here to the 97% (or 99.7%) solution?

-Jack (who does not speak enough languages to qualify for tackling
this problem, but loves the problem space)

Re: [9fans] combining characters

[quanstro]

that is precicely the most problematic case.  there is insufficient information in a bitmap font
to build an accented letter from a base codepoint + {combining codepoints}*.

- erik


On Fri May 19 16:28:05 CDT 2006, viriketo@gmail.com wrote:

> I've mostly used unicode diacritic mark "'" as in á, é, ś, ... (to be 
> composed over characters) specially over cyrillic characters, because 
> I'm learning Russian. There are no accented cyrillic letters in Unicode. 
> For a learner of the Russian language it's totally needed, if the main 
> source is written text. I hope that fits into that 97% of the future 
> (soon?) working solution.
> 

Re: [9fans] combining characters

[Lluís Batlle i Rossell]

[-- Attachment #1: Type: text/plain, Size: 907 bytes --]

Jack Johnson wrote:
> On 5/19/06, quanstro@quanstro.net <quanstro@quanstro.net> wrote:
> It sounds like there's definitely room for improvement, but I think
> it's a rabbit hole (with Alice in tow) to find the one general
> solution out there for the entire problem space.  Would it be great to
> find that solution?  Most definitely.  But that could take a lifetime
> or more of work.
> 
> So for me, the big question is what's the most graceful way to get
> from here to the 97% (or 99.7%) solution?

I've mostly used unicode diacritic mark "'" as in á, é, ś, ... (to be 
composed over characters) specially over cyrillic characters, because 
I'm learning Russian. There are no accented cyrillic letters in Unicode. 
For a learner of the Russian language it's totally needed, if the main 
source is written text. I hope that fits into that 97% of the future 
(soon?) working solution.

[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/x-pkcs7-signature, Size: 3311 bytes --]

Re: [9fans] combining characters

[jmk]

On Fri May 19 17:05:07 EDT 2006, knapjack@gmail.com wrote:
> ...
> So for me, the big question is what's the most graceful way to get
> from here to the 97% (or 99.7%) solution?
> ...

Try the 7% solution, it will make all your problems go away.

Re: [9fans] combining characters

[quanstro]

á is a single codepoint.  sure.  but there are useful letters that don't
exist in unicode unless they are composed.  e.g. romanized russian,
accented cyrillic, etc.

- erik

On Fri May 19 17:00:38 CDT 2006, nemo@lsub.org wrote:
> I think that á is just a single rune, not two different ones composed. If
> to type them, you have to type several keys, it´s just a keyboard issue,
> isn´t it? I don´t understand why this could go to a upper layer. Is there
> any other problem? (besides having to use utf8 for i/o, I mean).

Re: [9fans] combining characters

[Francisco J Ballesteros]

I think that á is just a single rune, not two different ones composed. If
to type them, you have to type several keys, it´s just a keyboard issue,
isn´t it? I don´t understand why this could go to a upper layer. Is there
any other problem? (besides having to use utf8 for i/o, I mean).


On 5/19/06, jmk@plan9.bell-labs.com <jmk@plan9.bell-labs.com> wrote:
> On Fri May 19 17:05:07 EDT 2006, knapjack@gmail.com wrote:
> > ...
> > So for me, the big question is what's the most graceful way to get
> > from here to the 97% (or 99.7%) solution?
> > ...
>
> Try the 7% solution, it will make all your problems go away.
>
>

Re: [9fans] combining characters

[Francisco J Ballesteros]

isn´t there enough space to keep all them there?

On 5/19/06, quanstro@quanstro.net <quanstro@quanstro.net> wrote:
> á is a single codepoint.  sure.  but there are useful letters that don't
> exist in unicode unless they are composed.  e.g. romanized russian,
> accented cyrillic, etc.
>
> - erik
>
> On Fri May 19 17:00:38 CDT 2006, nemo@lsub.org wrote:
> > I think that á is just a single rune, not two different ones composed. If
> > to type them, you have to type several keys, it´s just a keyboard issue,
> > isn´t it? I don´t understand why this could go to a upper layer. Is there
> > any other problem? (besides having to use utf8 for i/o, I mean).
>
>

Re: [9fans] combining characters

[quanstro]

i was really suprised by this, too.

i did a bit of work for a company that does searchable literature a couple
of months ago.  they were having trouble with "bad unicode".  the problem
 was stuff like this:

CA: Corporate Author
    Nizhegorodskai͡a͡ gosudarstvennai͡a͡
    selʹskokhozi͡a͡ĭstvennai͡a͡ akademii͡a͡


the character that probablly doesn't look right is a combining double breve.
it's actually good data.  i tracked down the cover of this book and it's really
spelled like that.

the problem is that the unicode folk didn't have the foresight to include
stuff like this.

- erik

On Fri May 19 17:05:24 CDT 2006, nemo@lsub.org wrote:
> isn´t there enough space to keep all them there?
> 
> On 5/19/06, quanstro@quanstro.net <quanstro@quanstro.net> wrote:
> > á is a single codepoint.  sure.  but there are useful letters that don't
> > exist in unicode unless they are composed.  e.g. romanized russian,
> > accented cyrillic, etc.
> >
> > - erik
> >
> > On Fri May 19 17:00:38 CDT 2006, nemo@lsub.org wrote:
> > > I think that á is just a single rune, not two different ones composed. If
> > > to type them, you have to type several keys, it´s just a keyboard issue,
> > > isn´t it? I don´t understand why this could go to a upper layer. Is there
> > > any other problem? (besides having to use utf8 for i/o, I mean).

Re: [9fans] combining characters

[quanstro]

On Fri May 19 17:30:53 CDT 2006, viriketo@gmail.com wrote:

> Francisco J Ballesteros wrote:
> > I think that á is just a single rune, not two different ones composed. If
> > to type them, you have to type several keys, it´s just a keyboard issue,
> > isn´t it? I don´t understand why this could go to a upper layer. Is there
> > any other problem? (besides having to use utf8 for i/o, I mean).
> 
> http://en.wikipedia.org/wiki/Combining_diacritical_mark
> 
> Think about languages having many diacritical marks (Arabic, Annamese, 
> Kannada,...), not only about Spanish. Wouldn't you feel there is too 
> much 'duplicated' information in those glyphs of that small 16bit space?

no.

Re: [9fans] combining characters

[Lluís Batlle i Rossell]

[-- Attachment #1: Type: text/plain, Size: 1268 bytes --]

Francisco J Ballesteros wrote:
> I think that á is just a single rune, not two different ones composed. If
> to type them, you have to type several keys, it´s just a keyboard issue,
> isn´t it? I don´t understand why this could go to a upper layer. Is there
> any other problem? (besides having to use utf8 for i/o, I mean).

http://en.wikipedia.org/wiki/Combining_diacritical_mark

Think about languages having many diacritical marks (Arabic, Annamese, 
Kannada,...), not only about Spanish. Wouldn't you feel there is too 
much 'duplicated' information in those glyphs of that small 16bit space?

Even more... there's the problem about _linking_ characters according to 
context, like in Arabic and at least some Indian languages. From Wikipedia:
"The Arabic alphabet can be encoded using several character sets, 
including ISO-8859-6 and Unicode, in the latter thanks to the "Arabic 
segment", entries U+0600 to U+06FF. However, neither of these sets 
indicate the form each character should take in context. It is left to 
the rendering engine to select the proper glyph to display for each 
character."

And after solving the screen output of unicode for diacritical marks and 
linking, maybe there'll be work to be done for the input.

[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/x-pkcs7-signature, Size: 3311 bytes --]

Re: [9fans] combining characters

[Roman Shaposhnick]

On Fri, May 19, 2006 at 05:16:22PM -0500, quanstro@quanstro.net wrote:
> i was really suprised by this, too.
> 
> i did a bit of work for a company that does searchable literature a couple
> of months ago.  they were having trouble with "bad unicode".  the problem
>  was stuff like this:
> 
> CA: Corporate Author
>     Nizhegorodskai?a? gosudarstvennai?a?
>     sel?skokhozi?a?i?stvennai?a? akademii?a?
> 
> the character that probablly doesn't look right is a combining double breve.

  What ?!?! How come this character ended up being a compound one ? The fact
  that it looks like one when rendered doesn't deprive it from being unique.

  Or am I missing something here ?

Thanks,
Roman.

Re: [9fans] combining characters

[quanstro]

i should have said "glyph."

the thing you're missing is that there is no precombined form
of {X+combining double breve} for any roman letter X that i know of.
this means that most utf renderers will give you either "X�" or �.

not having a precombined form is a pain.

- erik

On Fri May 19 17:35:43 CDT 2006, rvs@sun.com wrote:
> > 
> > CA: Corporate Author
> >     Nizhegorodskai?a? gosudarstvennai?a?
> >     sel?skokhozi?a?i?stvennai?a? akademii?a?
> > 
> > the character that probablly doesn't look right is a combining double breve.
> 
>   What ?!?! How come this character ended up being a compound one ? The fact
>   that it looks like one when rendered doesn't deprive it from being unique.
> 
>   Or am I missing something here ?
> 
> Thanks,
> Roman.

Re: [9fans] combining characters

[quanstro]

did you have a font with the precomposed characters on hand?
is that libpanel or libframe?

- erik

On Fri May 19 17:41:20 CDT 2006, mirtchovski@gmail.com wrote:
> we've discussed diacritics before:
> 
> http://pages.cpsc.ucalgary.ca/~mirtchov/screenshots/x-p9-diacritics.png
> 
> as with most other things, there wasn't enough interest to spark an
> implementation :)

Re: [9fans] combining characters

[andrey mirtchovski]

we've discussed diacritics before:

http://pages.cpsc.ucalgary.ca/~mirtchov/screenshots/x-p9-diacritics.png

as with most other things, there wasn't enough interest to spark an
implementation :)

Re: [9fans] combining characters

[Lluís Batlle i Rossell]

[-- Attachment #1: Type: text/plain, Size: 862 bytes --]

quanstro@quanstro.net wrote:
> On Fri May 19 17:30:53 CDT 2006, viriketo@gmail.com wrote:
> 
> 
>>Francisco J Ballesteros wrote:
>>
>>>I think that á is just a single rune, not two different ones composed. If
>>>to type them, you have to type several keys, it´s just a keyboard issue,
>>>isn´t it? I don´t understand why this could go to a upper layer. Is there
>>>any other problem? (besides having to use utf8 for i/o, I mean).
>>
>>http://en.wikipedia.org/wiki/Combining_diacritical_mark
>>
>>Think about languages having many diacritical marks (Arabic, Annamese, 
>>Kannada,...), not only about Spanish. Wouldn't you feel there is too 
>>much 'duplicated' information in those glyphs of that small 16bit space?
> 
> 
> no.
That called "unicode 4.0" comes with 96,248 graphic characters.
http://www.unicode.org/versions/Unicode4.0.0/

[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/x-pkcs7-signature, Size: 3311 bytes --]

Re: [9fans] combining characters

[quanstro]

i don't know hebrew very well, but are you confusing glyphs with characters?

for example arabic has three letter forms: initial, final and medial.
(there is a different shape for the the same letter at the beginning, middle
and end of the word.)

so in arabic, a good renderer would need three glyphs for each codepoint.

- erik

On Fri May 19 17:55:50 CDT 2006, joelcsalomon@gmail.com wrote:
> Take Hebrew, for instance: 27 letters (including the 5 final forms) +
> a few alternate forms, 15 vowel marks, 25+ cantillation marks --
> that's more than 10,000 combinations right there.
> 
> --Joel

Re: [9fans] combining characters

[Joel Salomon]

On 5/19/06, Francisco J Ballesteros <nemo@lsub.org> wrote:
> On 5/19/06, quanstro@quanstro.net <quanstro@quanstro.net> wrote:
> > á is a single codepoint.  sure.  but there are useful letters that don't
> > exist in unicode unless they are composed.  e.g. romanized russian,
> > accented cyrillic, etc.
>
> isn´t there enough space to keep all them there?

Take Hebrew, for instance: 27 letters (including the 5 final forms) +
a few alternate forms, 15 vowel marks, 25+ cantillation marks --
that's more than 10,000 combinations right there.

--Joel

Re: [9fans] combining characters

[andrey mirtchovski]

the top is X11 with proper diacritics composition, the bottom is Plan
9, without. in both cases the font is the same: Kuhn's 10646
(available as 'fixed' in plan9)

andrey

On 5/19/06, quanstro@quanstro.net <quanstro@quanstro.net> wrote:
> did you have a font with the precomposed characters on hand?
> is that libpanel or libframe?
>
> - erik

Re: [9fans] combining characters

[Roman Shaposhnick]

On Fri, May 19, 2006 at 05:35:10PM -0500, quanstro@quanstro.net wrote:
> the thing you're missing is that there is no precombined form
> of {X+combining double breve} for any roman letter X that i know of.
> this means that most utf renderers will give you either "X?" or ?.

  Well, but the language in question is Russian. In Russian, both of
  the symbols which can be seen as compound are unique symbols. 

  Of course, that only means, that whoever was entering the title
  into the system made a mistake in using wrong unicode.

Thanks,
Roman.

Re: [9fans] combining characters

[quanstro]

eh?  you speak russian. ;-)

no.  the unicode sequences (e.g. U+0069 U+0361) are correct.
i checked this and several other examples with the actual books.

i think you misunderstand how unicode works.  a base cp
like U+0069 followed by a combining cp like U+0361 
make a single character.  this identification is called "composition".
unicode contains some precomposed cps, but not U+0069 U+0361.

if you think this makes unicode bizarre, then join the club.  but
it does not mean that U+0069 U+0361 is an invalid unicode sequence
which represents a single character.

- erik

On Fri May 19 18:41:24 CDT 2006, rvs@sun.com wrote:
> On Fri, May 19, 2006 at 05:35:10PM -0500, quanstro@quanstro.net wrote:
> > the thing you're missing is that there is no precombined form
> > of {X+combining double breve} for any roman letter X that i know of.
> > this means that most utf renderers will give you either "X?" or ?.
> 
>   Well, but the language in question is Russian. In Russian, both of
>   the symbols which can be seen as compound are unique symbols. 
> 
>   Of course, that only means, that whoever was entering the title
>   into the system made a mistake in using wrong unicode.
> 
> Thanks,
> Roman.

Re: [9fans] combining characters

[Roman Shaposhnick]

On Fri, May 19, 2006 at 06:43:14PM -0500, quanstro@quanstro.net wrote:
> eh?  you speak russian. ;-)

  and two versions of it too ;-)

> no.  the unicode sequences (e.g. U+0069 U+0361) are correct.
> i checked this and several other examples with the actual books.

  How did you check it ? Visual inspection ? Since I'm no expert
  in UNICODE I'm quite curious to know how one is supposed to
  tell between a real character and a combination of a diacritic
  and some other character when they are visually indistinguishable ?
  I would expect unicode to always favor single glyphs from a particular 
  page over anything else.

  btw, could you send me a .png with the actual title ?

> i think you misunderstand how unicode works.  

  That could very well be the case ;-) But I know how Russian language
  works regardless of what committee members think.

> a base cp like U+0069 followed by a combining cp like U+0361 
> make a single character.  this identification is called "composition".
> unicode contains some precomposed cps, but not U+0069 U+0361.

  That's ok. My only point is -- I would expect anybody who enters 
  titles into a database adhere to the rules of the language the
  title is written in. Maybe its too much to expect, though.

Thanks,
Roman.

Re: [9fans] combining characters

[quanstro]

On Fri May 19 19:13:39 CDT 2006, rvs@sun.com wrote:
> > no.  the unicode sequences (e.g. U+0069 U+0361) are correct.
> > i checked this and several other examples with the actual books.
> 
>   How did you check it ? Visual inspection ? 

since these were actual books, i know of no other way. ;-)

>   Since I'm no expert
>   in UNICODE I'm quite curious to know how one is supposed to
>   tell between a real character and a combination of a diacritic
>   and some other character when they are visually indistinguishable ?

say i have a random accented letter.  suppose that U+x is the cp for
the letter.  suppose U+y is the cp for the accent.  suppose that we're lucky
and there exists U+w ≡ U+xU+y.  then U+w should be the same glyph
as U+xU+y.

cannonical composition would yield
	compose(U+xU+y)	U+w
	compose(U+w)		U+w
while cannonical decompostion would yield
	decompose(U+xU+y)	U+xU+y
	decompose(U+w)		U+xU+y


>   I would expect unicode to always favor single glyphs from a particular 
>   page over anything else.

it's always a single glyph.  don't confuse letters, codepoints, and glyphs.

> 
>   btw, could you send me a .png with the actual title ?

i'll send you a png of the character.  i don't have the books.

what language rule are you trying to get at?

- erik

> 
> > i think you misunderstand how unicode works.  
> 
>   That could very well be the case ;-) But I know how Russian language
>   works regardless of what committee members think.
> 
> > a base cp like U+0069 followed by a combining cp like U+0361 
> > make a single character.  this identification is called "composition".
> > unicode contains some precomposed cps, but not U+0069 U+0361.
> 
>   That's ok. My only point is -- I would expect anybody who enters 
>   titles into a database adhere to the rules of the language the
>   title is written in. Maybe its too much to expect, though.
> 
> Thanks,
> Roman.
> 

Re: [9fans] combining characters

[andrey mirtchovski]

> > i think you misunderstand how unicode works.
>
>   That could very well be the case ;-) But I know how Russian language
>   works regardless of what committee members think.

I know how the Bulgarian National Standard is set for the phonetic
keyboard (by the Bulgarian Academy of Sciences, no less) and that
still doesn't sway Apple: OSX misplaces several characters and even
puts Russian ones in (ы where ъ should be).

at least with Plan 9 I can edit the kbmap file if it was wrong (which it isn't).

Re: [9fans] combining characters

[Roman Shaposhnick]

On Fri, May 19, 2006 at 07:13:33PM -0500, quanstro@quanstro.net wrote:
> On Fri May 19 19:13:39 CDT 2006, rvs@sun.com wrote:
> >   Since I'm no expert
> >   in UNICODE I'm quite curious to know how one is supposed to
> >   tell between a real character and a combination of a diacritic
> >   and some other character when they are visually indistinguishable ?
> 
> say i have a random accented letter.  

  There's no such thing as an accented letter in a Russian language.
  That was the exact point of my initial remark.

  Now, if you allow me to educate myself in Unicode a little bit,
  I'm about to follow through with your example. Be patient with me ;-)

> suppose that U+x is the cp for the letter.  
> suppose U+y is the cp for the accent.

  Ok.

> suppose that we're lucky and there exists U+w ? U+xU+y.  

  Just to make sure I still follow: U+w is supposed to *visually* 
  look like U+x followed by U+y, right ?

> then U+w should be the same glyph as U+xU+y.

  The same glyph from a visual standpoint, right ?

> cannonical composition would yield
> 	compose(U+xU+y)	U+w
> 	compose(U+w)		U+w
> while cannonical decompostion would yield
> 	decompose(U+xU+y)	U+xU+y
> 	decompose(U+w)		U+xU+y

  And that's exactly the place where I think Unicode goes against common
  sense and language rules. I would expect it to mandate that a *decomposable*
  character is supposed to be used over the decomposition. Which in your
  original example was the case.

> >   I would expect unicode to always favor single glyphs from a particular 
> >   page over anything else.
> 
> it's always a single glyph.  don't confuse letters, codepoints, and glyphs.

  It is still a bit hard to not confuse letters and glyphs :-(

> i'll send you a png of the character.  i don't have the books.
> 
> what language rule are you trying to get at?

  "There are no accents in Russian language" (*)

Thanks,
Roman.

(*) well, except for a Ukrainian one ;-) 

Re: [9fans] combining characters

[quanstro]

On Fri May 19 19:45:43 CDT 2006, rvs@sun.com wrote:
>   There's no such thing as an accented letter in a Russian language.
>   That was the exact point of my initial remark.

the text was /romanized/ russian names.  it was not written in the cyrillic 
alphabet.

> 
>   Now, if you allow me to educate myself in Unicode a little bit,
>   I'm about to follow through with your example. Be patient with me ;-)

as long as you're patient with me.

> 
> > suppose that U+x is the cp for the letter.  
> > suppose U+y is the cp for the accent.
> 
>   Ok.
> 
> > suppose that we're lucky and there exists U+w ? U+xU+y.  
> 
>   Just to make sure I still follow: U+w is supposed to *visually* 
>   look like U+x followed by U+y, right ?

yes.  they must be the same.

> 
> > then U+w should be the same glyph as U+xU+y.
> 
>   The same glyph from a visual standpoint, right ?

a glyph IS the visual representation. 

> 
> > cannonical composition would yield
> > 	compose(U+xU+y)	U+w
> > 	compose(U+w)		U+w
> > while cannonical decompostion would yield
> > 	decompose(U+xU+y)	U+xU+y
> > 	decompose(U+w)		U+xU+y
> 
>   And that's exactly the place where I think Unicode goes against common
>   sense and language rules. I would expect it to mandate that a *decomposable*
>   character is supposed to be used over the decomposition. Which in your
>   original example was the case.

rob agrees with you.

however, there is a big advantage to a composed character -- you don't have to figure out
how to stick the horn, breve, slash, &c on top of, under, on the shoulder of, through, &c
the original character.  in plan 9, characters are bitmaps making this operation extra
annoying.  also, there are no rules in unicode preventing /arbitrary/ compositions.
this is valid unicode
	u+0069 u+0300 u+0301 u+0302 u+0303
all those combining codepoints attach to the base cp u+0069.  figure out how to build that
glyph.

> 
>   "There are no accents in Russian language" (*)
>

now you're confusing language and alphabet! ☺

- erik

Re: [9fans] combining characters

[quanstro]

that would be good to keep in mind. ;-)

- erik

On Fri May 19 20:00:34 CDT 2006, mirtchovski@gmail.com wrote:
> >   "There are no accents in Russian language" (*)
> 
> wikipedia disagrees:
> 
> Acute accents are also used in Slavic language dictionaries and
> textbooks to indicate lexical stress, placed over the vowel of the
> stressed syllable. This can also serve to disambiguate meaning (e.g.,
> in Russian писа́ть (pisát) means "to write", but пи́сать (písat) means
> "to piss").
> 
> http://en.wikipedia.org/wiki/Diacritic

Re: [9fans] combining characters

[andrey mirtchovski]

>   "There are no accents in Russian language" (*)

wikipedia disagrees:

Acute accents are also used in Slavic language dictionaries and
textbooks to indicate lexical stress, placed over the vowel of the
stressed syllable. This can also serve to disambiguate meaning (e.g.,
in Russian писа́ть (pisát) means "to write", but пи́сать (písat) means
"to piss").

http://en.wikipedia.org/wiki/Diacritic

Re: [9fans] combining characters

[Skip Tavakkolian]

for a description check the link below. pages 8 and 9 have
some of the detail.

http://behdad.org/download/Publications/persiancomputing/a007.pdf


> i don't know hebrew very well, but are you confusing glyphs with characters?
> 
> for example arabic has three letter forms: initial, final and medial.
> (there is a different shape for the the same letter at the beginning, middle
> and end of the word.)
> 
> so in arabic, a good renderer would need three glyphs for each codepoint.
> 
> - erik
> 
> On Fri May 19 17:55:50 CDT 2006, joelcsalomon@gmail.com wrote:
>> Take Hebrew, for instance: 27 letters (including the 5 final forms) +
>> a few alternate forms, 15 vowel marks, 25+ cantillation marks --
>> that's more than 10,000 combinations right there.

Re: [9fans] combining characters

[Roman Shaposhnick]

On Fri, May 19, 2006 at 06:59:31PM -0600, andrey mirtchovski wrote:
> >  "There are no accents in Russian language" (*)
> 
> wikipedia disagrees:
> 
> Acute accents are also used in Slavic language dictionaries and
> textbooks to indicate lexical stress, placed over the vowel of the
> stressed syllable. This can also serve to disambiguate meaning (e.g.,
> in Russian писа?ть (pis?t) means "to write", but пи?сать (p?sat) means
> "to piss").

  I don't think that wording is accurate. It gets close to the point
  though: "dictionaries and textbooks" are exactly the only place
  you might find these. But before I go on, I would like to ask
  our native English speakers: do you guys consider transcriptions
  used in the dictionaries a part of English language, a part of
  separate language or what ?

Thanks,
Roman.

Re: [9fans] combining characters

[Roman Shaposhnick]

On Fri, May 19, 2006 at 07:51:03PM -0500, quanstro@quanstro.net wrote:
> that would be good to keep in mind. ;-)

  For some reason it reminded me of 4 phrases every traveler should know:
      http://www.travelphrases.info

Thanks,
Roman.

Re: [9fans] combining characters

[Federico Benavento]

I'm not a native english speaker but phonetics is phonetics,
not a language, an alphabet.

http://www.arts.gla.ac.uk/IPA/images/ipachart.gif

On 5/19/06, Roman Shaposhnick <rvs@sun.com> wrote:
> On Fri, May 19, 2006 at 06:59:31PM -0600, andrey mirtchovski wrote:
> > >  "There are no accents in Russian language" (*)
> >
> > wikipedia disagrees:
> >
> > Acute accents are also used in Slavic language dictionaries and
> > textbooks to indicate lexical stress, placed over the vowel of the
> > stressed syllable. This can also serve to disambiguate meaning (e.g.,
> > in Russian писа?ть (pis?t) means "to write", but пи?сать (p?sat) means
> > "to piss").
>
>   I don't think that wording is accurate. It gets close to the point
>   though: "dictionaries and textbooks" are exactly the only place
>   you might find these. But before I go on, I would like to ask
>   our native English speakers: do you guys consider transcriptions
>   used in the dictionaries a part of English language, a part of
>   separate language or what ?
>
> Thanks,
> Roman.
>


-- 
Federico G. Benavento

Re: [9fans] combining characters

[geoff]

The pronunciation keys, at least in English dictionaries, use the
International Phonetic Alphabet, which is included in Unicode.

Re: [9fans] combining characters

[Jack Johnson]

On 5/19/06, quanstro@quanstro.net <quanstro@quanstro.net> wrote:
> a glyph IS the visual representation.

Hopefully.

http://en.wikipedia.org/wiki/Liber_Linteus

"There are 230 lines of text, with 1200 legible words. Black ink has
been used for the main text, and red ink for lines and diacritics."

Considering there are so few Etruscan texts left and that they may
never be translated, it would be a shame to lose this kind of
contextual information, especially if it happened to be useful in
deciphering the text.

-Jack

Re: [9fans] combining characters

[Lucio De Re]

On Fri, May 19, 2006 at 06:38:22PM -0700, Roman Shaposhnick wrote:
> 
>   I don't think that wording is accurate. It gets close to the point
>   though: "dictionaries and textbooks" are exactly the only place
>   you might find these. But before I go on, I would like to ask
>   our native English speakers: do you guys consider transcriptions
>   used in the dictionaries a part of English language, a part of
>   separate language or what ?
> 
The Italian-published Italian/English dictionary I occasionally
refer to represents the pronunciation of English words using a
phonetic alphabet.  The collection of those words is, in my opinion,
_not_ the English language, although I often wish it was because
I have a dyslexic girl-friend with an even more dyslexic son, they
battle valiantly with the written language with no hope ever to win.

Which shows that if the alphabet is simple, then some other factor
is brought in to complicate things, like totally inconsistent
spelling.

Go figure!  ISO are not unique.

++L

Re: [9fans] combining characters

[Lluís Batlle i Rossell]

[-- Attachment #1: Type: text/plain, Size: 180 bytes --]

Lucio De Re wrote:
> Which shows that if the alphabet is simple, then some other factor
> is brought in to complicate things, like totally inconsistent
> spelling.
Samopinie.

[-- Attachment #2: S/MIME Cryptographic Signature --]
[-- Type: application/x-pkcs7-signature, Size: 3311 bytes --]

Re: [9fans] combining characters

[Joel Salomon]

On 5/19/06, quanstro@quanstro.net <quanstro@quanstro.net> wrote:
> On Fri May 19 17:55:50 CDT 2006, joelcsalomon@gmail.com wrote:
> > Take Hebrew, for instance: 27 letters (including the 5 final forms) +
> > a few alternate forms, 15 vowel marks, 25+ cantillation marks --
> > that's more than 10,000 combinations right there.
>
> i don't know hebrew very well, but are you confusing glyphs with characters?
>
> for example arabic has three letter forms: initial, final and medial.
> (there is a different shape for the the same letter at the beginning, middle
> and end of the word.)
>
> so in arabic, a good renderer would need three glyphs for each codepoint.

Hebrew final forms are not much of a problem; they are separate
characters, typed with different keys.  It's the vowel marks needed
sometimes (whn yu cnnt nfr th vwls frm cntxt) and the cantillation
marks needed for Biblical text that make for the code space explosion.

Arabic text rendering—for readable plain text, not just for "fancy"
typesetting—requires yet another clever set of algorithms.

I don't know that there's a way to manage this complexity without
giving "fonts" their own context-aware programming language, like
TrueType.

--Joel

-- 
It reverses the normal flow of conversation.
> What's wrong with top-posting?
> > Top-posting.
> > > What's the biggest scourge on plain text email discussions?

Re: [9fans] combining characters

[Joel Salomon]

[-- Attachment #1: Type: text/plain, Size: 280 bytes --]

On 5/19/06, quanstro@quanstro.net <quanstro@quanstro.net> wrote:
> this is valid unicode
>         u+0069 u+0300 u+0301 u+0302 u+0303
> all those combining codepoints attach to the base cp u+0069.  figure out how to build
> that glyph.

The Gentium font makes a fair try.

--Joel

[-- Attachment #2: Nizhegorodskaia_gentium.png --]
[-- Type: image/png, Size: 29995 bytes --]

Re: [9fans] combining characters

[Joel Salomon]

[-- Attachment #1: Type: text/plain, Size: 223 bytes --]

On 5/19/06, Roman Shaposhnick <rvs@sun.com> wrote:
>   btw, could you send me a .png with the actual title ?

I've attached one, with four different "Unicode-capable" fonts.  I
suspect only the Arial gets it right.

--Joel

[-- Attachment #2: Nizhegorodskaia.png --]
[-- Type: image/png, Size: 68954 bytes --]

Re: [9fans] combining characters

[Joel Salomon]

[-- Attachment #1: Type: text/plain, Size: 280 bytes --]

On 5/19/06, quanstro@quanstro.net <quanstro@quanstro.net> wrote:
> this is valid unicode
>         u+0069 u+0300 u+0301 u+0302 u+0303
> all those combining codepoints attach to the base cp u+0069.  figure out how to build
> that glyph.

The Gentium font makes a fair try.

--Joel

[-- Attachment #2: Nizhegorodskaia_gentium.png --]
[-- Type: image/png, Size: 29995 bytes --]

Re: [9fans] combining characters

[Jack Johnson]

On 5/21/06, Joel Salomon <joelcsalomon@gmail.com> wrote:
> I don't know that there's a way to manage this complexity without
> giving "fonts" their own context-aware programming language, like
> TrueType.

What are the chances something like Display PostScript could've been
massaged to handle this?  Or is it really something to look at on a
font-by-font basis?

-Jack

Re: [9fans] combining characters

[Roman Shaposhnick]

On Sun, May 21, 2006 at 02:12:03PM -0400, Joel Salomon wrote:
> On 5/19/06, quanstro@quanstro.net <quanstro@quanstro.net> wrote:
> >this is valid unicode
> >        u+0069 u+0300 u+0301 u+0302 u+0303
> >all those combining codepoints attach to the base cp u+0069.  figure out 
> >how to build
> >that glyph.
> 
> The Gentium font makes a fair try.

  <horror>
  So that's how original looked like!
  </horror>

  The romanized spelling in the upper window is pretty convoluted. What
  they were trying to accomplish with that special "accent" is convey
  the idea of "ia" being a single letter in the original language. 
  And of course the letter itself doesn't have any "accents" in it. :-(

Thanks,
Roman.

P.S. Why would they use a second "accent" over the empty space between
two words still escapes me, though...

Re: [9fans] combining characters

[Roman Shaposhnick]

On Fri, May 19, 2006 at 10:59:37PM -0300, Federico Benavento wrote:
> I'm not a native english speaker but phonetics is phonetics,
> not a language, an alphabet.
> 
> http://www.arts.gla.ac.uk/IPA/images/ipachart.gif

  Fascinating... Thanks for the link!

Thanks,
Roman.

> 
> On 5/19/06, Roman Shaposhnick <rvs@sun.com> wrote:
> >On Fri, May 19, 2006 at 06:59:31PM -0600, andrey mirtchovski wrote:
> >> >  "There are no accents in Russian language" (*)
> >>
> >> wikipedia disagrees:
> >>
> >> Acute accents are also used in Slavic language dictionaries and
> >> textbooks to indicate lexical stress, placed over the vowel of the
> >> stressed syllable. This can also serve to disambiguate meaning (e.g.,
> >> in Russian писа?ть (pis?t) means "to write", but пи?сать (p?sat) means
> >> "to piss").
> >
> >  I don't think that wording is accurate. It gets close to the point
> >  though: "dictionaries and textbooks" are exactly the only place
> >  you might find these. But before I go on, I would like to ask
> >  our native English speakers: do you guys consider transcriptions
> >  used in the dictionaries a part of English language, a part of
> >  separate language or what ?
> >
> >Thanks,
> >Roman.
> >
> 
> 
> -- 
> Federico G. Benavento

Re: [9fans] combining characters

[Dan Cross]

On Fri May 19 19:45:43 CDT 2006, rvs@sun.com wrote:
>   There's no such thing as an accented letter in a Russian language.
>   That was the exact point of my initial remark.

This is true, at least for Cyrillics, but there are stress marks which, to
beginners in the language, are invaluable aids for sounding out the correct
pronunciation of words.  Typically, adult Russian isn't written with the
accent marks, though.  But children's books and textbooks for foreigners are.

	- Dan C.

Re: [9fans] combining characters

[Victor Nazarov]

Dan Cross wrote:

>On Fri May 19 19:45:43 CDT 2006, rvs@sun.com wrote:
>  
>
>>  There's no such thing as an accented letter in a Russian language.
>>  That was the exact point of my initial remark.
>>    
>>
>
>This is true, at least for Cyrillics, but there are stress marks which, to
>beginners in the language, are invaluable aids for sounding out the correct
>pronunciation of words.  Typically, adult Russian isn't written with the
>accent marks, though.  But children's books and textbooks for foreigners are.
>
>	- Dan C.
>  
>
Totaly agree with you. I'm a native russian speaker and I saw some books 
in russian wich use accents to overcome ambiguity and this usage seems 
elegant. But I need to say that accented texts seems very unusual at first.
--
Victor Nazarov

Re: [9fans] combining characters

[LiteStar numnums]

[-- Attachment #1: Type: text/plain, Size: 1752 bytes --]

These marks are in both of the Bulgarian texts I have, as well as the
Russian one.
They're extremely useful, esp. for Russian (the harder of the two in my
op.).

On 5/27/06, Victor Nazarov <vir@comtv.ru> wrote:
>
> Dan Cross wrote:
>
> >On Fri May 19 19:45:43 CDT 2006, rvs@sun.com wrote:
> >
> >
> >>  There's no such thing as an accented letter in a Russian language.
> >>  That was the exact point of my initial remark.
> >>
> >>
> >
> >This is true, at least for Cyrillics, but there are stress marks which,
> to
> >beginners in the language, are invaluable aids for sounding out the
> correct
> >pronunciation of words.  Typically, adult Russian isn't written with the
> >accent marks, though.  But children's books and textbooks for foreigners
> are.
> >
> >       - Dan C.
> >
> >
> Totaly agree with you. I'm a native russian speaker and I saw some books
> in russian wich use accents to overcome ambiguity and this usage seems
> elegant. But I need to say that accented texts seems very unusual at
> first.
> --
> Victor Nazarov
>
>


-- 
Nietzsche's first step is to accept what he knows. Atheism for him goes
without saying and is "contructive and
radical". Nietzsche's supreme vocation, so he says, is to provoke a kind of
crisis and a final decision about the
problem of atheism. The world continues on its course at random and there is
nothing final about it. Thus God
is useless, since He wants nothing in particular. If he wanted something --
and here we recognize the traditional
forumlation of the problem of evil -- He would have to assume responsiblity
for "a sum total of pain and inconsistency
which would debase the entire value of being born."
-- Albert Camus, L'Homme révolté

[-- Attachment #2: Type: text/html, Size: 2226 bytes --]

Re: [9fans] combining characters

[Andy Newman]

>   Speaking of which -- does anybody know a good technical paper on Quartz and
>   how it fits with the rest of the Darwin ?

This looks promising,

http://developer.apple.com/documentation/GraphicsImaging/Conceptual/drawingwithquartz2d/index.html

Re: [9fans] combining characters

[Roman Shaposhnick]

On Tue, May 23, 2006 at 01:07:57PM -0400, plan9@sigint.cs.purdue.edu wrote:
> Would be neat if someone cloned NeWS using Ghostscript as the engine, but
> Apple has already done Quartz with PDF

  Speaking of which -- does anybody know a good technical paper on Quartz and
  how it fits with the rest of the Darwin ?

Thanks,
Roman.

Re: [9fans] combining characters

[Harri Haataja]

On Tue, May 23, 2006 at 01:07:57PM -0400, plan9@sigint.cs.purdue.edu wrote:
> On Tue, May 23, 2006 at 10:24:14AM -0400, LiteStar numnums wrote:
> > On 5/23/06, erik quanstrom <quanstro@quanstro.net> wrote:
> >>
> >> DPS was the basis for OpenWindows and NeXTStep.
> >>
> > Wasn't NeWS 'doubly' PostScript: Display on one side, description
> > betwixt client & server?

> Would be neat if someone cloned NeWS using Ghostscript as the engine,
> but Apple has already done Quartz with PDF, so I suppose PostScript is
> passe'.  (Sorry, my old xterm doesn't do combined characters, just to
> jerk the thread back on track.)

Naturally an interested party re DPS was GNUStep.
http://gnustep.org/information/mission.html contains:
"
GNUstep has split the GUI into a front-end GUI "interface" and a backend
window-server specific implementation. With this architecture it is
possible to support several window-server backends (DPS, X, libart,
Windows). Our main interest is supporting the Display PostScript drawing
model (at least conceptually), but we may support other models in the
future.
"

I think they've only managed X11 though, but ICBW.

-- 
Being generous is inborn; being altruistic is a learned perversity.

Re: [9fans] combining characters

[Charles Forsyth]

> Would be neat if someone cloned NeWS using Ghostscript as the engine, but

i can't see combining two gargantuan peculiar characters as fitting under any of these definitions
(having excluded the cattle):

  Neat, a. [Compar. /Neater/ (?); superl. /Neatest/.] [OE. nett,
F. nett, fr. L. nitidus, fr. nitere to shine.  Cf. /Nitid/, /Net/,
a., /Natty/.] 

  1. Free from that which soils, defiles, or disorders; clean;
cleanly; tidy.
  If you were to see her, you would wonder what poor body it
was that was so surprisingly neat and clean. Law.
  
  2. Free from what is unbecoming, inappropriate, or tawdry;
simple and becoming; pleasing with simplicity; tasteful; chaste;
as, a neat style; a neat dress.
  
  3. Free from admixture or adulteration; good of its kind; as,
neat brandy. "Our old wine neat."  Chapman.
  
  4. Excellent in character, skill, or performance, etc.; nice;
finished; adroit; as, a neat design; a neat thief.
  
  5. With all deductions or allowances made; net. [In this sense
usually written net. See /Net/, a., 3.]
  neat line (Civil Engin.), a line to which work is to be built
or formed. -- 
  Neat work, work built or formed to neat lines.
  Syn. -- Nice; pure; cleanly; tidy; trim; spruce.

Re: [9fans] combining characters

[Ronald G Minnich]

plan9@sigint.cs.purdue.edu wrote:

> Actually, I think the client just sent DPS to the server, which rendered
> it to the screen.  I had hoped that Gosling would someday release the NeWS
> sources, but I imagine it's all tied up with Adobe IP.

don't forget the occasional stack vomit when things went badly. Ah, that 
was fun ...

fork() in postscript? make my day.

ron

Re: [9fans] combining characters

[plan9]

On Tue, May 23, 2006 at 10:24:14AM -0400, LiteStar numnums wrote:
> On 5/23/06, erik quanstrom <quanstro@quanstro.net> wrote:
>> On Mon May 22 23:59:33 CDT 2006, knapjack@gmail.com wrote:
>>> On 5/22/06, erik quanstrom <quanstro@quanstro.net> wrote:
>>>> it would work in the same way an ICBM could take care of a pesky
>>>> beetle.  effective, but not without colateral damage.
>>>
>>> I rescind my suggestion. For world peace. :)
>>>
>>> Actually, I had never taken a close look at OpenType.  I didn't
>>> realize just how much it can do.  Still a bazooka, but interesting
>>> anyway.
>>>
>>> -Jack
>>
>> DPS was the basis for OpenWindows and NeXTStep.
>>
>> - erik
>>
> Wasn't NeWS 'doubly' PostScript: Display on one side, description betwixt
> client & server?

It would have looked better in Doubly.

Actually, I think the client just sent DPS to the server, which rendered
it to the screen.  I had hoped that Gosling would someday release the NeWS
sources, but I imagine it's all tied up with Adobe IP.

Would be neat if someone cloned NeWS using Ghostscript as the engine, but
Apple has already done Quartz with PDF, so I suppose PostScript is passe'.
(Sorry, my old xterm doesn't do combined characters, just to jerk the
thread back on track.)

Re: [9fans] combining characters

[LiteStar numnums]

[-- Attachment #1: Type: text/plain, Size: 1391 bytes --]

Wasn't NeWS 'doubly' PostScript: Display on one side, description betwixt
client & server?

On 5/23/06, erik quanstrom <quanstro@quanstro.net> wrote:
>
> DPS was the basis for OpenWindows and NeXTStep.
>
> - erik
>
> On Mon May 22 23:59:33 CDT 2006, knapjack@gmail.com wrote:
> > On 5/22/06, erik quanstrom <quanstro@quanstro.net> wrote:
> > > it would work in the same way an ICBM could take care of a pesky
> > > beetle.  effective, but not without colateral damage.
> >
> > I rescind my suggestion. For world peace. :)
> >
> > Actually, I had never taken a close look at OpenType.  I didn't
> > realize just how much it can do.  Still a bazooka, but interesting
> > anyway.
> >
> > -Jack
>



-- 
Nietzsche's first step is to accept what he knows. Atheism for him goes
without saying and is "contructive and
radical". Nietzsche's supreme vocation, so he says, is to provoke a kind of
crisis and a final decision about the
problem of atheism. The world continues on its course at random and there is
nothing final about it. Thus God
is useless, since He wants nothing in particular. If he wanted something --
and here we recognize the traditional
forumlation of the problem of evil -- He would have to assume responsiblity
for "a sum total of pain and inconsistency
which would debase the entire value of being born."
-- Albert Camus, L'Homme révolté

[-- Attachment #2: Type: text/html, Size: 1886 bytes --]

Re: [9fans] combining characters

[erik quanstrom]

DPS was the basis for OpenWindows and NeXTStep.

- erik

On Mon May 22 23:59:33 CDT 2006, knapjack@gmail.com wrote:
> On 5/22/06, erik quanstrom <quanstro@quanstro.net> wrote:
> > it would work in the same way an ICBM could take care of a pesky
> > beetle.  effective, but not without colateral damage.
> 
> I rescind my suggestion. For world peace. :)
> 
> Actually, I had never taken a close look at OpenType.  I didn't
> realize just how much it can do.  Still a bazooka, but interesting
> anyway.
> 
> -Jack

Re: [9fans] combining characters

[Jack Johnson]

On 5/22/06, erik quanstrom <quanstro@quanstro.net> wrote:
> it would work in the same way an ICBM could take care of a pesky
> beetle.  effective, but not without colateral damage.

I rescind my suggestion. For world peace. :)

Actually, I had never taken a close look at OpenType.  I didn't
realize just how much it can do.  Still a bazooka, but interesting
anyway.

-Jack

Re: [9fans] combining characters

[erik quanstrom]

it would work in the same way an ICBM could take care of a pesky
beetle.  effective, but not without colateral damage.

- erik

On Sun May 21 23:05:27 CDT 2006, knapjack@gmail.com wrote:
> 
> What are the chances something like Display PostScript could've been
> massaged to handle this?  Or is it really something to look at on a
> font-by-font basis?
> 
> -Jack