Translations
============

(see TRANSLATIONS.txt for the Savane-specific options)

Generally, you create them using Django with either:

- a monolithic translation catalog
  (what we currently use)

  cd framework/
  mkdir locale/  # or conf/locale/
  django-admin makemessages -l fr
  django-admin compilemessages -l fr

- per-app translation catalogs:

  cd framework/savane/svmain/
  mkdir locale/
  django-admin makemessages -l fr
  django-admin compilemessages -l fr

- in both cases, to update:

  cd ...
  django-admin makemessages -a

Monolithic translations are not suited for deploying reusable apps
(because they are installed in the projects rather than in the
applications).

Per-app translation catalog means splitting the Savane 3 catalog.
It's not hard to do: just duplicate the current catalog in all
applications, and update the catalogs (all unused strings will be
commented out).

Per-app translation catalogs also means moving the templates in the
application directory, otherwise they won't be parsed.  For example:

  framework/savane/svmain/templates/svmain/group_detail.html

Also this means placing top-level templates in an application, so that
its strings get translated too (e.g. move templates/base.html to
framework/savane/svmain/templates/base.html).


Given that "reusable apps" in Django is mostly utopia as of 2010, and
given the constraints of per-app translation catalogs, let's use a
monolithic translation catalog for now.


Notes:

- Django needs to be restarted to take new translations into account
  (they are cached).

- Template {% blocktrans %} blocks takes newlines into account, so
  better write them on a signe line, even if it's very long.

- {% blocktrans %} uses mappings ('%(variable)s' instead of '%s'),
  which is not compatible with Savane 3's .po strings.  Consequently
  we'll get a lot fuzzy strings that translators will need to review.
  Thus let's not worry so much about introducing fuzzy strings:
  typically let's not keep All Firstcap Titles that cripple Savane 3.

- The strings discover parser is limited and doesn't understand
  "string1" + "string2" on multiple lines, so don't be afraid to use
  super-long text lines in the code.