Comments on: Mozilla DTD files, caveat emptor

By: Axel Hecht

Axel Hecht — Thu, 05 Feb 2009 01:06:15 +0000

barf. I always thought that mailnews code was bad, and now you made me look at bugzilla. how evil.

By: Vitaly Fedrushkov (SnowyOwl)

Vitaly Fedrushkov (SnowyOwl) — Wed, 04 Feb 2009 22:50:10 +0000

Dwayne: ergonomic UI is always concise. There are web pages and error messages which make things complex. You have asked about real-life example, consider this one:

http://mxr.mozilla.org/mozilla/source/webtools/bugzilla/template/en/default/global/user-error.html.tmpl#111

This code is designed to emit more than 60 similar error messages and is still correct in English. Do you want to persuade developers that having 60 separate sentences is better ‘from l12y standpoint’?

I’m already engaged in uphill battle of bug 407752.

By: seth bindernagel

seth bindernagel — Wed, 04 Feb 2009 00:51:16 +0000

@ Dwayne: the point of l20n is not to make things harder. We want to make things better. Good point though.

By: Dwayne Bailey

Dwayne Bailey — Tue, 03 Feb 2009 11:22:16 +0000

The summary for me is that when we try be fancy we end up having to be fancier to work around the problem we’ve created.

I’d be curious to see some examples beyond the Firefox one, issues such what to do with types of data in variables that are more closely aligned to what programs output: filesnames, website titles, etc

I’m reviewing some LOCALIZATION NOTES and found some of these:
Blah + (a,b or c). Where a, b and c are on average 3 word string. It seemed like an amazing solution to a coder yet it would simply have worked better to have 3 redundant strings:

Blah a
Blah b
Blah c

And let localisers and there tools manage the differences.

One thing I would like to echo which Friedel mentioned, is that if l20n makes things harder for the many languages that do not have this problem or don’t need to manage it yet. Then I’m afraid we fail.

By: palmiche

palmiche — Sat, 31 Jan 2009 02:08:03 +0000

It’s been a while but I think Sid Meier’s 4Xs from Colonization to Alpha Centauri have a very interesting l10n system design. I’m no expert so try abandonia.com for a copy of Colinization and take a look at the language file, it might (hopefully) provide a different perspective and ideas for improvements.

By: Vitaly Fedrushkov (SnowyOwl)

Vitaly Fedrushkov (SnowyOwl) — Fri, 30 Jan 2009 23:37:34 +0000

F Wolff: “So the interesting thing is that localisers in other projects struggle less with this since the need for rebrandability is not so strong elsewhere.”

In Bugzilla, the term for principal database object ‘Bug’ is customizable. Imagine Firefox has user-configurable terms for ‘bookmark’ and ‘tab’ and all UI texts must be changed accordingly and adapt to nouns of any gender.

By: Nicholas Shanks

Nicholas Shanks — Fri, 30 Jan 2009 11:34:18 +0000

Another major headache when translating things is plural forms.

In English, the singular is used for integer 1, the plural for all other integers.
In French, the singular is used for integers 0 and 1, and the plural for all other integers.
In Polish, there are different forms for integer ranges 1, 2-4 and 5-21.
Arabic has six plural forms.

See http://doc.trolltech.com/qq/qq19-plurals.html for a great summary. The pootle page http://translate.sourceforge.net/wiki/l10n/pluralforms has a list of equations, and it in turn uses GNU’s gettext function to achieve this, http://www.gnu.org/software/automake/manual/gettext/Plural-forms.html

By: Staś Małolepszy

Staś Małolepszy — Fri, 30 Jan 2009 10:48:10 +0000

@ Goofy, F Wolff:
Rebranding is one reason you may want to do this, but it is really a side effect to something much more bigger: thinking about Mozilla as a platform. If you hard code “Firefoksa” everywhere, even Thunderbird and SeaMonkey can no longer share the same l10n code. Making sure we use &brandShortName; all the time is in fact encouraging developers to experiment with Mozilla code and create new applications (browsers or not) with less l10n effort.

By: F Wolff

F Wolff — Fri, 30 Jan 2009 10:09:51 +0000

Merike, I don’t believe that the choice has to be between simplicity and correctness. KDE’s Transcript is entirely hidden for languages that don’t need to use it, but provides the power to those who need it. I would say that simplicity is an absolute requirement for any alternative solution. Things are already hard. Making it harder won’t benefit our cause.

By: Merike

Merike — Thu, 29 Jan 2009 20:37:05 +0000

In addition to branding and beforementioned Eastern languages’ differences localization also gets tricky with dates, one example being bug #463273 where code rewriting introduced difficulties with localizing.
As far as I understand it’s a choice between simplicity for languages that don’t have declensions and making good localizations possible for those that do. It just can’t be comfortable for both at the same time.
Anxious to see what will be the solution that you mention in your next post. To me it seems that for use cases where one string is put into another there could be a base form of the string where localizer can omit the ending and then add correct ending in the resulting string. But that requires a base case that isn’t used separately.