[haiku] Re: Haiku's localization vocabulary guidelines

Vincent Duvert <vincent.duvert@xxxxxxx> wrote:
 ...
> Annuler is already used for both "Undo" and "Cancel". (*)
 ...
> (*) And that's why you should not make dialogs like this
> : "Unable to undo this operation. [ Cancel ]". 
> Once translated, this gets pretty confusing :-D

Nice example. :]

(warning: long piece on my own ideas on translation)

I remember thinking once when I was like six years old
that English must be like Swedish (my mother tounge)
except you replaced each character by another one - 
simple character substitution, and hey presto, English!

The world at large seems to be stuck on the idea that
language can be subdivided in words or phrases and that
these can be individually translated and then reassembled
with no ill effect, to produce natural and good language.

I think it's a bad idea to assume that basing translation
on lookup of phrases from a primary language to secondary
languages has no ill effect, or even that is is the most
economical way to do mass-language translation. Seeing 
that one must annotate the primary language's phrases
with context and so expose every language's more exotic
nuances to the primary language and thus to all other 
target languages, to allow for contextual variations of
phrases in all languages that need them.

Think: X languages * Y phrases * Z contexts
 => a lot of phrase variations
 => a lot of accumulated complexity.

But I think there is a way to sidestep this.

It may seem economical to translate all occurances
of a phrase (e.g. "Send") in the same application or
in all applications on a system. I bet the 
administrative overhead of contextualizing a phrase 
initially thought to be simple and cascading this
change to all translations severly limits the
speed/progress of localization.

Instead of pushing the context -into- the lookup key
and trying to share this phrase+context key among all
translations, cut both context and master phrase out of
the lookup. Treat all phrases as unique occurences.
Do all translations -in context-, per application, per
phrase occurence, per string instance. Have the lookup
be a unique identifier (say a number) of this phrase
instance, not tied to any master language, but as a 
lookup key to any translation.

E.g. application "Adam" (e-mail) has three occurences
of the phrase "Send", in its user interface. Each of 
these occurences would use a sufficiently unique(*)
lookup number in the source code:

2980435454353223 -> "Send", "Skicka", "Utli", ...
1986254565658899 -> "Send", "Skicka", "Utle", ...
7891794565468387 -> "Send", "Skicka", "Utla", ...

(* Some number range community authority might be
helpful. Application could be assigned a "big" number
to be added to the local phrase key small number, and
when combined they'd be a universal phrase occurence 
pointer of sorts. All could be looked up, and cached
or embedded.)

This sets the stage for languages deviations,
giving languages more independence from one another.

When every string instance is independently looked
up, there's no technical need for a phrase change in
a certain context to consider the large amount of other
contexts in which the phrase occurs.

Phrase consistency across applications and systems,
stops being enforced at the lower level and becomes
a higher-level question of aestetics and taste. As it
should be.

This would allow translators to only pay attention to
producing the most natural language for each phrase
occurance, within its context, without worrying about
every other occurance of the same or similar phrases.

-Being able to translate- separate occurences of a 
phrase differently without the forethought of the
original developer and the expressiveness of the 
master language, is good thing.

This does not mean a developer/translator would have
to translate the same phrase -over and over-. Think
RAD tools and collaboration websites.

In my mind the ideal is to have a sea of translations
with good dynamics and flow from any one language to 
any other, not limited by translators' command of a 
master language (usually English) or lack thereof.

Coherency between translation should come about not
by hiearchy but by flow and cross-pollination.

If I were to translate, I would like very much to be
presented (by a website or application) not only an
English phrase and some notes on context, but a few
related but differing phrases in all of the languages
that I know somewhat.

/Jonas.

Other related posts: