🚀 go-pugleaf

[Note move of thread from soc.culture.nordic to comp.std.internat, as well
 as a single hint to readers of soc.culture.turkish in case of interest.]

In article <TT.92Sep9114439@tarzan.jyu.fi>, tt@tarzan.jyu.fi (Tapani Tarvainen) writes:
> In article <1992Sep8.160511.1976@corax.udac.uu.se> andersa@Riga.DoCS.UU.SE (Anders Andersson) writes:
> >The Turkish alphabet is different here (and more consistent, in my
> >opinion), as it has two different vowels 'i'; one with dot and the
> >other without (both letters of course appear in upper- and lowercase).
> >I'm afraid even ISO 10646 fails to support them properly...
> 
> I'm fairly sure even some less ambitious ISO character set
> (probably 8859-n, where n>1) supports Turkish completely, including
> the dotless i (treated as a separate character).

I suggest we look at ISO 8859-3 (which I understand is the official name
for Latin Alphabet Nr 3) for reference, as it claims to support Turkish.
In the following, I'm disputing the 'completeness' of that support:

Latin-3 contains among other, mostly southern European, characters

  0xA9  capital letter I with dot above, and
  0xB9  small letter i without dot above.

Of course, these are supposed to be used in conjunction with the
'normal' ASCII characters of the LH part of the table, in particular

  0x49  (Latin) capital letter I, and
  0x69  (Latin) small letter i,

to make up the two different kinds of 'i' used in Turkish, each in
upper- and lowercase.  From a mere typographic standpoint (having a
unique code for each visually distinguishable glyph), I consider
this support complete.

Programmers are used to being able to perform case conversion on
letters of the ASCII table by simply adding or subtracting a certain
constant to the character code, given that the code is within a
particular range (A-Z or a-z).  With later ISO standards, this is
not quite such a simple task due to the sometimes ad-hoc layout of
lowercase letters with respect to corresponding uppercase letters
(examples available upon request), but it would still be possible
using tables showing the relationship.

However, since the same character code is now used for both Latin
capital 'I' and Turkish capital dotless 'I', case conversion is no
longer a trivial matter.  Consider TO_LOWER(TO_UPPER(dotless 'i')).
It ought to be symmetric, but what's the result?

Is it somehow understood that automatic case conversion of letters
of the Latin, Greek and Cyrillic alphabets (and possibly others) is
beyond the scope of ISO character standards, or is this just an odd
case having been overlooked?  Judging from the little I've seen of
ISO 10646, it contains no better support for Turkish 'i' variants
than Latin-3 does (see positions 0x0130 and 0x0131 in UCS-2).

My proposal: Add two specifically Turkish letters to ISO 10646,
one capital 'I' without dot and one small 'i' with dot, and consider
them different from the Latin 'I' and 'i'.  I have no formal
relationship with any standardization body, so I'll have to leave
this proposal for any interested party to bring it up in the proper
forum.

Latin and Cyrillic capital 'M' look the same, while the small forms
don't.  Those capital 'M' letters have different codes in ISO 10646,
though maybe for reasons of systematic tabulation rather than in order
to support case conversion.  We did away with the old typewriter
unification of '1' and 'l' long ago (and the same for '0' and 'O',
if that ever was a problem).  Is Turkish 'i' vs. Latin 'i' in that
different a ballpark?

Are there word processors today that know how to case-convert a word
containing Turkish letters?  What are Turkish typists used to?

Are there other letters in other alphabets suffering from similar
unification problems in current ISO standards?
--
Anders Andersson, Dept. of Computer Systems, Uppsala University
Paper Mail: Box 325, S-751 05 UPPSALA, Sweden
Phone: +46 18 183170   EMail: andersa@DoCS.UU.SE