Created attachment 108395 [details] [review]
make actions' names and descriptions to be translated

Created attachment 108460 [details] [review]
modified patch--make actions' names and descriptions to be translated

I am sorry that there is something improper in the last patch, and I have modified it this time.

I just wrote in bug 351477 that I'm leaning towards marking action names for translation as a potentially bad thing to do.  Instead, I'm thinking maybe we should unmark the strings in Evolution and keep action names as untranslated strings.  Thoughts?

Does the user ever see the strings that are the action names?
If not, why do they need to be translated? I agree with Will
and would prefer that they weren't translated. This is going 
to cause a lot of headaches getting it right across numerous 
locales.

Any progress on figuring this one out ? 

My naive assumption would be:

action names: untranslated
descriptions: translated

 

(In reply to comment #5)
> My naive assumption would be:
> 
> action names: untranslated
> descriptions: translated

This would probably be the most expedient thing to do and would be more in line with there things are today.  The only change needed would be to "untranslate" the strings marked for translation in Evolution.

Li, what are your thoughts?

(In reply to comment #6)
> (In reply to comment #5)
> > My naive assumption would be:
> > 
> > action names: untranslated
> > descriptions: translated
> This would probably be the most expedient thing to do and would be more in line
> with there things are today.  The only change needed would be to "untranslate"
> the strings marked for translation in Evolution.
> Li, what are your thoughts?

It is a good idea. The only problem is Gok only show actions' names. We may need Gok shows descriptions of actions also.

(In reply to comment #7)
> (In reply to comment #6)
> > (In reply to comment #5)
> > > My naive assumption would be:
> > > 
> > > action names: untranslated
> > > descriptions: translated
> > This would probably be the most expedient thing to do and would be more in line
> > with there things are today.  The only change needed would be to "untranslate"
> > the strings marked for translation in Evolution.
> > Li, what are your thoughts?
> 
> It is a good idea. The only problem is Gok only show actions' names. We may
> need Gok shows descriptions of actions also.

David, what are your thoughts?  The main issue is that Evolution is currently the only one to flag accessible action names for translation.  In Orca, we currently look at the action name in order to infer information about the object (e.g., a "toggle" action on a table cell helps us infer it might be a checkbox).  As such, the current GAIL implementation of not flagging the action names for translation works well for us, but the Evolution implementation causes things to fail for other languages.  

Comment #5 seems to provide a reasonable middle ground that will keep things kind of at the status quo if we end up un-marking action names in Evolution.  The question for you, however, is what impact does this have on GOK?

So what's the current policy on accessibility vs. localization? What parts of the ATK API should be filled with english texts and what parts should be filled with translated texts? How should automated tools (e.g. test suites) discover elements and what texts should tools for users (e.g. screen readers) present to the user?

IMHO, the conclusion to this bug is easy:

atk_action_get_name  => untranslated
atk_action_get_localized_name  => translated
atk_action_get_description => translated

So the current patch are wrong, and instead of returning a translated action name on get_name, it should implement atk_action_get_localized_name (after a brief look, gtk is not implementing that).

At the beginning for me was hard to understand why the discussion took so long, as atk_action_get_localized_name is available since 2002. So I took a look to atspi and it lacks the localized_name option. So probably all the people involved were thinking in at-spi terms, and thought that the localized_name option was not available. I will add a bug on atspi about that.

(In reply to comment #9)
> So what's the current policy on accessibility vs. localization? What parts of
> the ATK API should be filled with english texts and what parts should be filled
> with translated texts? How should automated tools (e.g. test suites) discover
> elements and what texts should tools for users (e.g. screen readers) present to
> the user?

The general rule is that any string that potentially could be shown to the user should be translated. Action names are a corner case because some could implement some kind of heuristic based on the name. But as I explained this was solved by providing atk__action_get_localized_name

(In reply to comment #10)
> IMHO, the conclusion to this bug is easy:
> 
> atk_action_get_name  => untranslated
> atk_action_get_localized_name  => translated
> atk_action_get_description => translated
> 
> So the current patch are wrong, and instead of returning a translated action
> name on get_name, it should implement atk_action_get_localized_name (after a
> brief look, gtk is not implementing that).

Question for gtk developers: should I change the bug summary to something like "atk_action_get_localized_name is not implemented"?

> At the beginning for me was hard to understand why the discussion took so long,
> as atk_action_get_localized_name is available since 2002. So I took a look to
> atspi and it lacks the localized_name option. So probably all the people
> involved were thinking in at-spi terms, and thought that the localized_name
> option was not available. I will add a bug on atspi about that.

Done: bug 680598