Open Stack

Hi I18n team,

I just push my patch which changes horizon to generate the language
list automatically [1]
and the this topic is traced by the bug [2].

I would like to share my long-standing idea with the team.
Does it sound reasonable?

* The basic idea is to generate the language list based on message
catalogs (PO files) availability.
* Horizon now has four types of message catalogs (django, djangojs) in
horizon and openstack_dashboard respectively.
  If all message catalogs are present, such languages are listed in
the language list.

There are some corner case logic.

* Some languages share a same locale prefix.
  Good example is Spanish variants: es, es-ar, es-mx and es-ve.
  When horizon searches message catalogs for es-ar, a locale "es_AR"
is first checked
  and then "es" (prefix) locale is checked. In this case, if "es-ar"
has no corresponding
  translations ("es_AR") and "es" is already registered in the
language list, "es-ar" will
  not be added in the language list.
* For the source language (i.e., English), message catalog existence
check is skipped.
* The current setting LANGUAGES still has effect. Operators can use e
used to limit
   the initial set of languages to exclude languages they want to show
to their users.
   (Note that even when a language is specified in LANGUAGES setting,
    if not all message catalogs for the language exist the language
will be skipped.)

Thanks,
Akihiro

[1] https://review.openstack.org/#/c/450126/
[2] https://bugs.launchpad.net/horizon/+bug/1675298