[josm-dev] JOSM Translations

[Dirk Stöcker]

On Sun, 2 Oct 2011, Frederik Ramm wrote:

> With JOSM continuously gaining more features, maintaining full translations, 
> especially those spoken only by a small group of mappers or worse, only by 
> translators who are not even familiar with mapping, becomes harder.

When there is not a single change in translation for months then degradion 
of quality is to be expected. The languages I mentioned have not even been 
near 100% in their best days in the past.

> For this reason - but not for this reason alone - I would love to see 
> something like a "JOSM light", a version of JOSM that brings the benefits of 
> a stable, standalone, offline-capable editor without tons and tons of extra 
> features.

I know I know, but I don't think this will happen :-) But we plan an 
expert/novice mode to make it easier for beginners.

> I have no idea how this should be achieved. One way would certainly be to 
> reverse course and put more stuff back into plugins - possibly improving the 
> plugin architecture along the way. Trying to fork off a JOSM light would also 
> be an option.

I don't see any features of core which aren't really important parts and 
could be stripped. Maybe some of the tools functions could, but this would 
neither change software size nor translation count in a noticeable way.

> This would enable us to have an almost fully translated "JOSM light" that 
> beginner users could use without having to fear the occasional un-translated 
> English message.

>From my experiences with other software developed in the past I know, that 
the amount of texts is not really relevant for the fact whether a 
translation is complete or not. More important is if a person feels 
responsible or not.

That texts change is normal part of software development. I rather have a 
partially untranslated JOSM than an unchanging one, as that would mean the 
software is dead.

The original mail is mainly a trial to find someone responsible for the 
abondoned languages. Other projects are much more strict when dropping 
languages (they have limits like 90%, not 20%).

> At the same time, there would be clear priorities for translators - translate 
> the JOSM light messages first, then progress to the more complex things 
> needed for the full feature set. At the moment, a translator would not know 
> if "You moved more than {0} elements." is more or less important to translate 
> than "Please launch the preferences dialog and retrieve another OAuth token." 
> ...

In Launchpad the texts are already ordered in the order of importance: 
First core strings, then presets and additional texts, then plugins.

> Even if we fail to make a proper "JOSM light", that last issue - knowing 
> which texts are part of everyday mapping and which aren't - is something that 
> we could investigate. Would it, for example, be possible to use different 
> translation wrappers in the code - e.g. tr1(), tr2(), tr3() instead of only 
> tr() - to denote importance?

Hardly. While this is possible in general it wont work. I.e. error texts 
are usually never visible, but when they are, you need to understand them 
or you are totally lost. Most of the strings in the software are thus 
strings, which are seldom or never shown, but when they are they are 
important.

What is possible already today is to add a comment in front of tr(), 
giving hints to the translators. I usually add I18N as keyword in the 
text, like this: /* I18N: Don't translate the text in brackets */

Currently we have approx. 50 to 150 strings per release, which means a 
workload of 5 strings a day. If a translator uses 5 hours each month, he 
could probably translate about 300-500 strings and a year later one of the 
bad languages could reach 100%. For Chinese I have seen that JOSM complete 
can be done in two days.

Ciao
-- 
http://www.dstoecker.eu/ (PGP key available)