International texinfo

[Rodrigo Medina]

                Dear Sirs:

Some years ago I tried to write a manual in Spanish using Texinfo. I gave
up.
There were too many problems. Recently I tried again, having heard that
there
has been an effort in internationalizing Texinfo. I was disappointed, the
main problems are still there. In English Texinfo is terrific, in almost any
other language it is just terrible.  This time I tried a little harder and
I was able to overcome some of the difficulties. Here I give my suggestions
of
how Texinfo should behave in order to be useful in non-English documents.

1- General  comment
  The initial strategy, of having a unique, pure ASCII, source document is
the
  correct one. Any effort in allowing the .texi document to be written in
  extended alphabets is a step in the wrong direction. The efforts shuold
  concentrate into obtaining the right output.

2- Problems with Tex output.
 a- A mistake.
    The first problem is just a mistake, and it should be easy to solve. The
   letters "i" and "j" with accents above them should be constructed with
the
   versions of such letters without the dot. The acute_i IS "\'\i", IT IS 
   WRONG to use "\'i" (Knuth dixit). The wrong output that texinfo yields
   is really awful, it does not correspond to quality that a TeX document
   usually has.  On the other hand, as far as I know, @dotless{i} and
   @dotless{j} are useless.

 b- A perhaps more difficult problem is the fact that cross-resferences do
not
   function properly when the node names have accents and other macros.
   To prohibit the use of accented letters in names of chapters etc. is
   unacceptable. It is like pretending that in English that chapter names
   shuold not contain, say, the letter "q". Unfortunately  I am not expert
   enough of TeX insides to give an exact solution. I can give two
suggestions.
   First: to delay the accent translation until the cross-reference job is
   finished.
   Second: Maybe it is easier to differentiate between a node_name and a
   node_title. The node_name should be used for the cross-references, and
   should not contain any macros; the node_title would be what is actually 
   written in the output and could contain accents, macros etc. If
   node_title is not defined  it should default to node_name.

  c- A suggestion: To give a macro for as many as possible signs that have
  equivalents in the IsoLantinX series of symbols.


3- Problems with makeinfo.
   The principal problem with makeinfo is the poor quality of what it yields
   for the accents and symbols. It is unacceptable for today's standards
that
   in an instalation with, say, IsoLatin1 the acute_a be written as "a'".
What
   I have done in the past was to go over the whole doccument changing the
   accents, one by one. That obviously is not a solution. What I think
makeinfo
   should do is the following. The doccument shuold have an instruction
telling
   makeinfo to what IsoLatinN code it wants to be translated. Makeinfo
   should yield the proper IsoLatinN code for all the accents and symbols
   that are defined for such IsoLatin table. The other symbols should be
   rendered as they are now.

        If these problems are solved, I think  that texinfo could be used
very successfully with non-English languages.

4- In the mean time.
   In order to obtain IsoLatin1 symbols from makeinfo I have done the
following.
   I have made a new source file (.txim) which is identical to the usual
.texi
   file except from the fact that all IsoLatin1 accents that apear are
written
   with 2 @, instead of 1, e.g. acute_a is @@'a instead of @'a. The literal
@
   should apear as @@@@ and the literal } as @@@} instead of @@ and @}
   respectively. This file is processed with makeinfo, that yields a .info
file
   in which the cross refereces are OK and all isolatin1 accents are now in
   the form they whuold had in a .texi file. The only @ remaining in the
file
   are precisely those accents. Now the file is processed with a simple
program
   (infoiso1) that translates the accents codes to the IsoLatin1 number.
That
   gives a perfectly working .info file, with the isolatin1 accents even in
   the name of chapters and so on.
        In order to obtain the TeX output the .txim file is processed with
   a program (mtotxi) that eliminates the double @, yielding a standard
  .texi that can be then processed by TeX.
        The whole procedure functions quite well, except from the problems
of
   texinfo we have told above.


I hope that the above comments could be useful in order to obtain a working
internationalized Texinfo.

Rodrigo Medina address@hidden

mtotxi.c
Description: mtotxi.c

infoiso1.c
Description: infoiso1.c