[XeTeX] Using tipa.sty with XeLaTeX

[Arthur Reutenauer]

> So surely one thing that should be done is to let the TIPA active-Z
> be defined as something else, say  \textezh , which should give  U+02D1 .

  You mean U+0292.

> This frees up  \textyogh  for its *correct* meaning, and still gives
> TIPA users the correct character when using shorthands.

  Yes.

> This is what  xunicode  was meant to do, but I forgot to define   
> \textezh .
> Good thing, in a sense, else this discussion would not be taking  
> place. :-)

  Interestingly enough, you *also* had to give TIPA active-2 an
incorrect definition (\textezh instead of \textturnv) for the error to
show up in the particular example Chandra sent ;-)

> If \textyogh is already used directly in existing TIPA documents,
> then the desired glyph can be recovered using:
> 
>     \AtBeginDocument{\let\textyogh\textezh}
> 
> perhaps given as the result of a package-option [yogh] to   
> xunicode.sty  .

  I like this solution a lot.  Maybe you could include all spurious
characters in this option and call it [tipa] since people compiling old
documents with the \letter... commands would probably want to have the
whole bunch of (faulty) character names defined at the same time.

  But in any case, this way of action seems the right direction to me:
we want to support old documents, but we don't want that to be a
hindrance for future usages (and anyway documents using the character
commands should not be the majority).  You could probably even call that
option by a name clearly indicating that it's meant for already existing
documents like [tipacompat] for example.

> (Such documents cannot want  \textyogh  for both glyphs!)

  This would not be reasonable.  But since I've met a student of
linguistics who inputted every single language in her (LaTeX) document
with a different scheme, I do not expect reasonable anymore :-) (she had
to type text in 3 different scripts and 5 or 6 different languages, and
she inputted Greek with Babel transliteration; Lituanian with a lot of
accent commands; Russian with direct Unicode encoding because someone
had convinced her into it; of course you had to add Sanskrit, a bit of
Latvian, as well as the main language of the document, French).

> and there are some macros defined in  tipa.sty  where I just didn't know
> what was the correct letter;  e.g.
>      is  \textlhtlongi   U+0196  or  U+01AA  or something else ?

  I would say that it is the same Unicode character as \textlhti (U+027F
LATIN SMALL LETTER REVERSED R WITH FISHHOOK), and that it should
actually be preferred over the former for representing that character
(as far as I understand from http://std.dkuug.dk/JTC1/SC2/WG2/docs/n2366.pdf
and the TIPA manual).  This character has been devised by a Swedish
linguist for transcribing Swedish dialects, and somehow it came to be
used in Chinese linguistics; whereas the Japanese (in books about
Chinese linguistics) use a slighly shorter variant glyph, \textlhti.
This doesn't help, I know.

  But maybe the right solution is to map \texthtlongi to U+027F and
leave \textlhti “unencoded” as far as Unicode is concerned; people would
have to input it with the TeX command (after all, there is no TIPA
shorthand for it).  The point is the glyph-to-Unicode application is not
one-to-one anyway, so people cannot expect a particular glyph for
representing a Unicode character; or they have to use “higher-level
protocols”, as the Unicode Standard puts it.  One of those higher-level
protocols can be the use of TeX commands.

> Hope this helps,

  Sure, it's very promising.

	Arthur