Translate.org.za Development Blog

Translate.org.za recently decided to work more closely with the upstream OpenOffice.org localisation process. In many ways this is partly because OpenOffice.org will be adopting Pootle to allow online translation for all localisation projects. But also because we’d like people going to the OpenOffice.org site to know about the Afrikaans translations.

As a result of this we now have our own Afrikaans project website which is directly accessible from the multilingual OpenOffice.org download page.

We are also now part of the official build process for localized efforts of OpenOffice.org and the first Afrikaans build already appeared in one of the recent 2.3 milestones.

The purpose of these builds are to allow testing of the localizations, so feel free to install and test this. We value your feedback.

We are still committed to providing our multilingual builds, but will either work a way to integrate those into the current process or build them as needed following official releases.

I’m usually suspect of machine translation. Its gets most things wrong in my opinion, but then again it is useful for getting the gist of a paragraph.

Andreas has been working with people from Apertium, doing MT form English to Afrikaans and back. Hard but then also not hard in that Afrikaans and English are related. You can give the it a whirl in their test environment. How about trying “Ek is ‘n boer”, you’ll notice its actually very good at Inglish

We’re excited about this as the potential for close languages like the Nguni and Sotho groups of languages is very good.

As you might or might not know the international dialing prefix for outgoing international calls changed from 09 to 00 recently. You didn’t know? Try dialing 09 now and see how far you get.

To make sure our glibc locale data is up to date we released a new version of the locales. Actually its just English that needs to change as all the other 10 inherit the telephone section from it.

What does this mean? Not much, except that applications that get the international select code from the locale will now get South Africa’s correct. For us its also a simple indicator that free software really is empowering. We’ve changed this we haven’t had to wait for work arounds while someone feeds this up to the powers that be.

You can download the new locale data package (you need to know what you’re doing) or track the glibc bug.

When we added the extra Venda characters, ḓṱḽṋṅ, to the DejaVu font we were newbies.  Needless to say the diacritics didn’t all line up correctly.  Ouch.  This looked really bad when viewed in large sizes when doing a demo.

At a recent conference Dwayne met with Denis Jacquerye, a DejaVu font contributor.  He volunteered to fix the problem and those fixes appeared in r1626 and r1627.  These should appear in DejaVu 2.16, which is scheduled to be released on 1 April.

What does this mean for you?  We’re not sure ourselves!  We use the orthography files from fontconfig in a small tool used by the Deja Vu font team that can analyse a font to arrive at a percentage coverage for various languages. With this tool we can quickly check if a font covers Afrikaans, Venda, Tswana and Northern Sotho.

Our guess is that font-config itself uses these orthography files in its algorithm for font substitution.  For instance if a font does not have Venda characters it can find a font that does cover the Venda characters and use that for the missing glyphs.

In our bug report we added some missing South African languages and fixed the entry for Tswana which didn’t include the Šš characters.

We’ve just submitted a bug to GNOME which adds character pickers for Afrikaans and a seperate one for Venda, Tswana and Northern Sotho.

What does this mean? Character Picker is a GNOME applet, its sits on the GNOME taskbar and displays a small list of buttons.  Useful if your keyboard doesn’t have those characters and you’re not sure how to type them.  In South Africa there are a number of people who need to type a few characters who find the keyboard difficult to understand or don’t want to mess with their keyboard.

Of course we have a keyboard!  But this is just a nice to have little extra. We certainly will find it useful as although we have keyboards setup sometime we simply need the list of characters easily accesible.

Just got an email from Matthias Klose, developer for Ubuntu and packager of the Ubuntu OpenOffice.org packages.  He’s looking at our spellcheckers for inclusion in Ubuntu.  We’re solving a few build problems with him but hopefully soon we’ll see all our checkers on a Linux computer near you.  Thanks Matthias.

We’ve just completed the migration of the translations at pootle.translate.org.za from CVS to Subversion. This now completes our migration off of the translate.sourceforge.net project into our own zaf.sourceforge.net project. This should make it easier to get translations contributed by the community back into Subversion at ZAF.

We have begun building OpenOffice.org 2.0.4 on Linux, Windows will follow as soon as we have a stable build environment. Don’t be confused with our 2.0.3 test builds which are ironing out build issues on the last stable release. The 2.0.4 work is to get a version building cleanly, once we have done that we will test that all Translate.org.za related mods are included and onpy then will 2.0.4 replease our 2.0.3 versions.

We will publish builds when they are available for testing.

CIA, the commit tracker as apposed to the US inteligence agency, is now tracking commits for South African translations. All seems to be working but I might need to adjust some things later.