Re: GNOME Localization and Usability

[Danilo Segan <danilo gnome org>]

Hi Reinout,

Reinout van Schouwen <reinout@cs.vu.nl> writes:

> http://www.cs.vu.nl/~reinout/pdf/GNOME_localization_and_usability.pdf

I didn't finish reading yet, but I've got a remark to make on one issue.

In the "(In)accessibility" section, when you talk about using keyboard
shortcuts in menus/controls, you miss important features of Gtk+
which help alleviate this problem (which plagues English as well).

By simply pressing the shortcut key twice (or more times), you may
cycle through all the items with the same shortcut keys. So, in any
case, it's substantially better to provide shortcut keys for each and
every element of the UI, rather than neglect it because of fear of
causing a collision. 

This also means that it's not neccessary to cycle with the Tab key
through all the widgets, because it usually takes at most two (it's
really rare that translator reuses same shortcut key more than twice) 
Alt+<mnemonic> or simply <mnemonic> (in menus) to reach desired
destination.

This does hinder the usability, but not to the extent you seem to
imply in your article.


As one of the more active translators of GTP, I must also add to this
that some of the problems you mention are also present in original
English strings. That's what we usually call "bugs", and that's what
we report.  The advantage English has over all other languages is
that it's proofread by every translator (you can probably notice that
most of the typos in bugzilla are reported by translators, not users
of English language).  Number of users also plays a role, since
default POSIX locale is English-based, and some users are not aware
how to change the locale to their own language.

This number of proofreaders is what distinguishes quality of English
strings, because there's no chance that any other language has
between 20 and 40 (my guesstimate of average number of translations
per module) *careful* readers of *all* the strings. Most of the
strings are seen in rare occasions, and it's sometimes not even
possible to really test them, because translator's computer doesn't
provide what it takes for a certain dialog to appear.

So, my opinion is that only a number of readers and users would make
a big difference.

On the consistency part, this has so far been delegated to
translation teams.  I'm setting up software that will be usable as a
web glossary, where everyone will be able to add suggestions for
translation, and translation team coordinator will be able to mark
one (or more) suggestions as "official translation".  Other teams
have done the same, perhaps in a different form.

Also, as you point out yourself, there exists no biective (1 to 1, and
each word has a corresponding word in the other language) mapping
between natural languages, so it's impossible to set any artificial
limits on how should translators translate what.

Overall, I agree with most of your points, and active participation
of translators is what also helps solve those problems
(eg. translators should report bugs to bugzilla when context is not
clear, so developers would add "translator comments" which would show
up in PO files next to the questionable string).

Cheers,
Danilo