Блогчетање

Већина вероватно не зна, али од издања 0.15 алата GNU gettext, укључен је и алат recode-sr-latin.

Алат је веома брз (писан је са тим циљем у C-у, а онда га је Бруно Хаибле [Bruno Haible: изговор?] још додатно оптимизовао), а ево примера за ПО фајл из Еволуције (~700kb):

$ time recode-sr-latin <sr.po >sr@latin.po

real    0m0.072s
user    0m0.056s
sys     0m0.008s
$ time cir2lat <sr.po >sr@latin.po

real    0m0.649s
user    0m0.412s
sys     0m0.160s

(cir2lat је моја стара скрипта са скоро еквивалентном функционалношћу, тј. ради и оно што описујем испод).

Додатно, recode-sr-latin ради и боље од већине претварача из српске ћирилице у латиницу, а ту мислим на исправан рад (колико је то могуће програмски постићи) у претварању речи као што су „Његош“ и „ЊЕГОШ“ у латиницу:

$ echo "ЊЕГОШ није Његош" | recode-sr-latin
NJEGOŠ nije Njegoš

Пошто данас већина система садржи геттекст 0.15, који је изашао у јулу 2006. време је и да се објави ова новина ;)

Some very cool things happening for local Gnome users. A certain Serbian company, has become a Novell partner for Serbian market. What they're interested in is getting a fully localised system to be able to offer it to government institutions and everybody else. Novell is willing to sponsor getting such a system.

They somehow got to me, and asked me if I'd like to help with that (to try to actually involve everybody who's already into it in Serbia). Of course, I was very much interested.

The catch is that they want a fully capable small business system, and are doing this with next release of SuSE in mind (of course, they care about enterprise uses, so NLD is also in consideration, but SuSE, due in September, is the first step): which means that the work will need to be carried out until end of July. Unfortunately, I was not able to find anything about what software is included in next SuSE release (I tried developer.novell.com and googling around), and I was quick to realise why is closed development not such a great thing.

Since they're letting me estimate the workload and see if this is doable, you can guess that "supported" Serbian desktop is going to be based on Gnome, and I'm very strongly considering pushing Gnome Office as a default office package. If only Criawips was actually useful (hey Herzi, I promise to resolve bug #172459 by end of June so you can go ahead and work on Criawips :), Gnome Office would definitely be a default (even with AbiWord translation problems I mentioned before: I'd simply fork the translation to make it consistent). This way, it depends on our ability to actually finish up OpenOffice.org translation in time.

There are many things that need to be done, but they're so hard to do with their process being so closed. This includes things like fonts (they're still buggy in everything except Debian I think), keyboard layouts, locales, spell-checking with aspell and OOo, PDF viewing (things like search should "Just Work" for Serbian texts, as far as technically possible [I know that one can make completely unsearchable texts as PDFs], but I doubt that evince is coming there), proof-reading all translations, documentation, etc.

So, if you have already worked in Serbian l10n field for free software systems, feel free to contact me at 063 8626 082 or danilo@gnome.org by this Thursday evening. I'll contact several people I have in mind myself, but don't let that stop you helping us and helping yourself!

Just to give kudos to the new Fedora Serbian Translation team.

They already rock very hard, and I provided them with some webspace and mailing list.

Also, I turned comments off in my blog because of the big amount of spam (I don't want to spend my time on installing CAPTCHA and similar tests).

Mark: not using Gnome CVS is a big problem for translation, if I may say. I tried contacting Serbian AbiWord translator earlier, but never really got through.

Not surprisingly, AbiWord looks completely out of place on an otherwise fully Serbian Gnome desktop: it's written using Latin script instead of Cyrillic, it uses completely different menu names (though they're recognizable and often quite similar, they're still different!), and I simply can't get to learn enough about simply getting involved and I lack the time to chase Serbian AbiWord translator for too long.

AbiWord doesn't really fit: notice a pattern in other menus?

FWIW, this is not only a problem of AbiWord: Christian recently raised the same issue with Inkscape developers (yeah, they do have good practical reasons not to switch to Gnome CVS, but it always means a step back for translation users and contributors).

Best help you can get from Gnome Translation People (ok, it's officially "Project", not "People" :) is by using Gnome CVS!

Disclaimer: this doesn't mean that I don't appreciate what you're doing for helping AbiWord be as close as possible to what translators otherwise do, like using PO files, creating statistics such as this, etc. It's just that it's never good enough. Of course, feel free to ask me for help regarding those stats generation, I have done quite a bit on that front myself :)

Убунту полако постаје најбоља дистрибуција ГНУ/Линукс система за српски језик: већ су укључени исправљени фонтови (грешка #2117; ово је исправљено и за наредно издање Дебијана), а ускоро (надам се) ће бити укључени и исправни локалитети за српски језик (грешка #5604). Све ово очекујте у наредном издању „Hoary“ које нас очекује за неких недељу дана.

Сад још само да неко почне да допуњава преводе (има врло мало посла да се уради, пошто је читав Гном 2.10 преведен, а њега користи Убунту), и моћи ћемо да препоручујемо Убунту свима (а још када не би због ограничења места избацили српске преводе са инсталационог и „уживо“ диска, где би нам био крај ;)!

I've found some time to hack a bit more on my PO handling stuff: I've added PO file merging along with "inline-diffs", along with an example pomerge.py utility.

As for diff-format, I've used what GNU wdiff seems to use to mark out changes (I'm getting redirection limits exceeded on these pages now, perhaps they're down; you can look at any of the mirrors, such as here).

At the same time, this wouldn't be so useful if our PO editors didn't add some support for it: well, you can see that I have po-mode with syntax highligthing (and there's also a polite-mode which has nothing but syntax highlighting, thus is useful for testing).

Emacs highlighting diffs for fuzzy matches.

Now, why do we need this? Look at the examples above: our second diff clearly indicates that the only change for a matched message was removal of "\n" at the end of line. First diff shows that one word was removed, other was added, and newline was removed. And with all this, it's so much easier to see what needs to be changed in translation.

Granted, with small messages such as those in the screenshot, this isn't very useful. But, when it comes to translating documentation (hint, hint: xml2po), this should be a very major help when someone fixes a typo or adds some interpunction or only a couple of words to a long paragraph.

Now, with Python this was easy: who feels like working on something like this for GNU msgmerge?

Roozbeh: thanks—finally, after almost two years of trying, sr_CS locale is in GNU libc! :)