Re: different encodings for input and output file names and command line

[Gavin Smith]

On Mon, Feb 28, 2022 at 10:29:02AM +0100, Patrice Dumas wrote:
> Hello,
> 
> First, some tests with locales different from the @documentencoding made
> me realize that it would make sense to have a different encoding of
> file names for output than for input.  Indeed, it may make sense for the
> input files names (@include, @verbatiminclude) to match the document encoding
> when extracted from an archive, for example.  But even in that case
> encoding the output names using the @documentencoding is very dubious,
> especially since we often use a different output encoding for the file
> content than the @documentencoding.
> 
> Another use case stirs even more in this direction.  In the init files
> (latex2html, tex4ht, highlight syntax), there are some commands launched
> from texi2any.  It seems to me natural and less error prone to encode
> those command lines in the locales encoding.  But this also forces the
> file names referenced in those command lines to be in the locales
> encoding.
> 
> In any case, handling all the different situations makes it necessary
> to be able to disconnect the input file names encoding from the output
> file names encodings, consistently with what we do with documents
> encodings.  I will do the corresponding code anyway and allow both the
> locale to be used and the @documentencoding in both cases based on
> customization variables switches.  The question that remains is what to
> use for the defaults and for the command lines.
> 
> My proposal is:
> 
> * input file name encoding:
>    My preference would be the locale, but Gavin proposal to use
>    @documentencoding also has merit, so let stick to @documentencoding
>    except on Windows where the locale is used.
> * output file encoding:
>    Use the locale in the default case.
> * command lines called from texi2any
>    Always use the encoding already used for messages defaulting to the
>    locale encoding
> 
> Opinions, ideas?

Yes, I agree with all of this.  For creating files it would make sense for
the locale encoding to be used or for UTF-8 where the output format requires
this (as you said in your other mail).