[bug-gnu-libiconv] iconv history

[Bruno Haible]

Info for the recode maintainer:

Reuben Thomas wrote:
> See https://sourceware.org/bugzilla/show_bug.cgi?id=29913

When iconv was designed (HP-UX had the first implementation of it, IIRC),
the common charsets were mostly variations of US-ASCII or EBCDIC. So,
when iconv, before 1995, by default mapped
   La recherche a échouée.
   -> La recherche a ?chou?e.
or
   Die Datei konnte nicht geöffnet werden.
   -> Die Datei konnte nicht ge?ffnet werden.
it made some sense, because the result was still human-intelligible.

Then, while glibc's iconv was developed (around 1998-2001), it became
clear that it also had to handle Japanese and Chinese character sets,
and mapping
   无法打开文件。
   -> ???????
was completely useless.

That's why //TRANSLIT is not the default in glibc or GNU libiconv.
An application cannot just throw strings like "??????? ???? ??????"
to the user's face. It has to find alternatives. And iconv() refusing
to produce such utterly useless conversion results is part of the
solution.

Bruno