[Localization] Indigenous language L10n processes

Chris Leonard cjlhomeaddress at gmail.com
Sun Jul 24 01:51:59 EDT 2011

Dear Localization Community,

There are a variety of methods for facilitating third language
localization via an intermediate (non-English) language.  This is an
important process for South American indigenous languages via Spanish
(Aymara, Quechua, Nahuatl, Tzotzil, Huastec / Tének, etc.) as well as
Francophone Africa and other circumstances where a non-English
majority lingua franca overlaps an indigenous language.

== Option 1 ==

The Translate Toolkit has a tool called poswap which can process a
translated PO file to replace the msgid with the mstgstr.  Starting
with a completed Spanish PO file, you get Spanish strings in msgid.


== Option 2 ==

The folks at Runasimipi are working to translate software into Quechua:


and they've developed a tool called instrans that can process a
translated PO file to include the msgstr as a translator comment).
Starting with a completed Spanish PO file, you get English strings in
the msgid with Spanish strings (previously in the msgstr) as a
translator comment.


Please examine the product of these different processes as they appear
on Pootle.

poswap version (abacus)

instrans version (abacus)

Both versions should be fully functional in off-line editing
situations with a tool like Virtaal (and important consideration).
I'm less familiar with poedit or other off-line tools though, testing
is encouraged.

>From a Pootle management point-of-view, the instrans processed files
are probably easier to manage because they can be hosted in Pootle
with commit links established and commits performed by the language
administrator as usual, With poswap, I'm not sure that Pootle would
properly handle commits where the msgid does not match the original
POT and manual intervention would probably be needed to reverse the
poswap process and return the msgids to English for commit.

Ease of management is one element to weigh, but ease of localization
is obviously the critical piece of the puzzle.

Do you think that instructing the indigenous localizers to pay
attention to the translator comments would be adequate cueing for
successful Spanish to Huastec L10n (as one example) on the instrans
processed PO files?

I look forward ot hearing your thoughts on this.  In an ideal world,
we would adopt a single standard process for addressing these
situations when they arise rather than having a mix of solutions for
different languages, so a broad range of feedback is desirable as are
alternatives to the one's I've proposed.  I suspect some
experimentation will be required to fully explore these options (e.g.
to fully understand how each works with gettext plurals, pre-existing
translator comments and other edge cases).

Warmest Regards,

volunteer Sugar Labs / OLPC / eToys Poolte administrator

More information about the Localization mailing list