#============================================================================

# Enca v1.19 (2016-09-05)  guess and convert encoding of text files

# Copyright (C) 2000-2003 David Necas (Yeti) <yeti@physics.muni.cz>

# Copyright (C) 2009-2016 Michal Cihar <michal@cihar.com>

#============================================================================

TO THE NEXT RELEASE:

(this list must be empty at the time of release)

IN FUTURE:

(should be done, but maybe not right now)

* LCUC check for cyrillic charsets.

* Backups -- like cp, mv, etc.  This will be hard to get right with all the

  silly converters.

* More tests

* Structured documentation (the manual page is ugly)

  - keep a reasonably brief manual page

  - put all the boring doc stuff somewhere else, there are possibilities:

    info: searchable, has links, partly portable, has console viewers

    HTML: poorly searchable, has links, most portable, has console viewers

    TeX (ps): not searchable, no links, portable, most pleasant to read,

          no console viewers

    => use SGML (or info itself?) and generate the others

MAYBE SOMEDAY:

(when I will have mood for it, items are freely moved here and removed again)

* Detect all-caps texts OK.

  After several experiments it seems we have to

  - use pair occurences, at least, with specificaly computed

    difference-maximising weights

  - guess in two steps

  - first with uncapitalization and pair weights, and check whether the

    sample looks like natural text (garbageness test, but better)

  - if the first approach fails, do it as we do it now

* design better levels of verbosity/warnings (or: remove the --verbose option,

  keep important messages and remove all others?)

  0: only messages followed by exit(EXIT_FAILURE) (or abort()) are printed

     plus `cannot convert...'

  1: all nonfatal errors/warnings

  2: what converters are tried, what language gets detected (do not duplicate

     --details)

  >2: debug

* _real_ paranoiac behaviour assuring that nothing gets lost and that

  conversion output is either correctly converted text or untouched original

  (requires major redesign of all the conversion stuff)

NEVER:

(you can do anything GNU GPL v2 allows, but I'll restrain)

* features that nobody needs (mm, well, ... ok, let it be)

* duplicate other tools functionality more than necessary, use them instead

* dependency on anything that is not ISO C and/or POSIX (moreover do not use

  braindead features of both); important functionallity must be present

  everywhere nevertheless, enca can be smaller, faster or cleverer on some

  (GNU) systems

* localization; please correct my english instead ;->

* converter calling generalization (would require inlcuding the whole wordexp

  thing in enca, and: launching external converter is Bad Thing(TM) anyway)

* data in run-time files (needs parser (could live with) and disallows hooks

  (can't live without))

* loadable module support (it's not very portable)

-------------

KNOWN ISO C CONFLICTS:

(perhaps to be solved someday)

All constants and typedefs.  They start with ENCA_ and Enca, but:

  Names beginning with a capital `E' followed a digit or uppercase

  letter may be used for additional error code names.               [errno.h]

And additionally inside libenca (i.e. not so serious):

* libenca.h: #define EPSILON                                        [errno.h]

* filters.c: isvbox[]                                               [ctype.h]

* guess.c: #define isbinary                                         [ctype.h]

* guess.c: #define istext                                           [ctype.h]

* multibyte.c: is_valid_utf7()                                      [ctype.h]

* multibyte.c: is_valid_utf8()                                      [ctype.h]

Some probably can't conflict.