rosetta language variant support

Arc Riley arcriley at
Tue Oct 6 17:04:03 BST 2009

> What do you mean with 'language variants'?  Language codes using
'xx_XX at something' are not supported fully.

just xx at something for the translation, though xx_XX at something would be the

> You can use 'make suggestions from <language>' to show another language
> translations as
> suggestions (if it's not a language using above mentioned language code
> style).

That's a neat feature, but I'm really just looking for being able to plug
the data in right now.  I assume that teams managing translations have the
ability to edit the files directly and thus upload them in batch rather than
string by string through rosetta?

In most cases (transliterations) it's a just matter of mapping words, not
phrases.  Most variants I'm aware of have the same spoken version as their
base, just a change of written form so the word ordering should remain the
same.  We already have scripts for automating much of this process though I
see no reason why they couldn't eventually be integrated with rosetta.

For the language variant I'm interested in we already have a sqlite database
with over 120k words transliterated and checked over by hand.  There's only
a few words, mostly names, that need to be transliterated by hand or
otherwise corrected.

> Launchpad is open source (get started at, so
> if you are interested in getting help on integrating a feature like this
> into Launchpad, we'd be happy to help.  Though, note that it'd involve a
> lot of basic infrastructure work to accommodate future variant support
> as well.

At the moment, I'd really just like to get going on this and be able to
integrate our translations normally rather than build our own
"transliteration pack" package, which seems messy and a lot of unnecessary

Can rosetta currently just treat the language label as unique instead of
trying to group it with it's base language?  IE, instead of treating
"sr at latin" as related to "sr" in any way, just accept the string "sr at latin"
as a separate translation matched by the exact string?

There are several translation groups for variants that are currently unable
to work within Ubuntu because of this, while I agree being able to show the
"sr" text against the "sr at latin" text would be helpful in the rosetta UI, it
doesn't seem right to prevent a sr at latin team from doing their work in lack
of this feature.

This would also get us around any problems of rosetta not allowing "en" as
our base language, "en at shaw" might as well be "fr" if the strings are
matched directly.

In our case we already have<>replacing
the "@" with "-" as is done in XML where @ isn't a valid
-------------- next part --------------
An HTML attachment was scrubbed...

More information about the ubuntu-translators mailing list