[PAL-en] RE: Au sujet de la localisation de Site Internet

["Don Osborn" <dzo@xxxxxxxxxxxx>]

-------> [ENGLISH] AUTOMATIC TRANSLATION NOT REVISED FROM THE ORIGINAL IN french

[ answer of Stephan Bortzmeyer on the list "Unicode-.Afrique" *;
retransmitted with small changes to facilitate the translation ]


I submitted the question to the expert "localization" at Debian
[ Christian Perrier ] and here his answer (approximately, the problem
exists, but does not have a simple solution).


> I thus tested the method GETTEXT which seems to be more standard. This
> technique makes it possible to store the translations in files PO
> which are much more convivial to handle by the translators. The
> technique functions with wonder on my servor localhost (WAMP5: we work
> under Windows), but when I transfer the files on the Web server from
> my hebergor, nothing goes any more!
> 

Already, there it is an excellent idea.: -)


> 1. Initially, it seems that the language (locale) must be preset on
> the servor of the hebergor. If not, the order setlocale of GETTEXT
> does not function. Since the national languages of Senegal are not
> defined at my hebergor then only English and French function!
> 

I am not really sure that there are locales for all the
languages of Senegal. Perhaps under Windows this is the case... with
the glibc it has there only one local for Wolof. 

In any case, if the hebergor does not have the locale it is sure that
that will have difficulty in go.


> 2. Then, the original chains of our site are in French, and not in
> English. They are encodées in UTF-8 and not in US-ASCII. Therefore,
> in the files PO, the chain msgid are encodées in UTF-8. On the
> servor, only the msgid not containing accents (éèçëà, etc.) are
> translated: we thus have a translation partial of our English site.
> Handbook GETTEXT indicates to us that the msgid must obligatorily be
> encodés in US-ASCII! It is a little ridiculous for a system which
> aims at internationalization!
> 

In fact, gettext (recent versions) functions very well with
original chaînes in UTF-8. But it is true that it was never made for
other thing but of the English translation - Another language. 

I think that it could be interesting to check if gettext is not a
little old on the servor. 

Example: the file fr.po which I attach to this mall. The recent
utilities gettext will not detect it as invalid although it contains
UTF-8 on some chaînes.


> I would like also to invite you to reflect on the following problems:
> 
>  * In general, one always translates out of English. But if one must
> translate out of another language because our translators do not
> control English? It is often an obstacle met in French-speaking, or
> Portuguese-speaking Africa.
> 

It is a problem known in the community of localization, yes
(one could quote the same problem for the native languages of the
Latin America -- the translator quechua will prefer to work starting
from Spanish).

Several projects in progress currently work on the possible mapping in
this field:

- to have a file complete French PO

- to make use of it, in parallel with the "English" POT to
create a file POT where the msgid are in French.

- to have the translator wolof who produces a wo-fr.po based above

- remapper in the other direction to make a "traditional"
wo.po

I wonder even if one could not do that with the tools gettext,
in fact....

But, there are projects as Pootle which have plans to work
thereabove... just probably not the arms [ people to do it ]. It is an
interesting subject for a programmer of free software, in fact.



*
http://fr.groups.yahoo.com/group/Unicode-Afrique/message/1256 

[ Notes of translation 
.chaîne = string
.hébergeur = web hosting service