Re: [boost] [Locale] Boost.Locale - preliminary code

Subject: Re: [boost] [Locale] Boost.Locale - preliminary code
From: Artyom (artyomtnk_at_[hidden])
Date: 2009-11-09 10:29:17

Hello,

> 
> I just took a quick peek at the docs, and noticed that
> message translation was based on GNU gettext. IIUC, the
> gettext library's license is LGPL, and its translation tools
> are GPL. AINAL, but I'm concerned that this may be
> incompatible with the Boost License. Can you comment on
> this?
> 

Yes, fist of take a look on Design Rationale:
http://cppcms.sourceforge.net/boost_locale/docs/index.html#design-rationale

Section: "Why to use GNU Gettext catalogs for message formatting?"

In short: 

1. Boost.Locale **does not** use GNU Gettext runtime at all. it
   reimplements the runtime and solves one of big gettext problems
   thread-safety with multiple locales.
   
   So there is no licensing issue at all.

2. Another issue is support of plural forms that so far is
   the best and most mature one in Gettext.

   ICU supports plural forms only starting from release 4.0
   and plural forms are hard-coded in the locale definition
   rather then defind as flexible as in Gettext.

3. ICU still does not allow working directly with XLIFF file
   formath but rather you should convert XLIFF to ICU resource
   boundles.

4. Availability of free tools for locaization.

So I find GNU Gettext more reliable choise for this purpose,
also it is much more widely used, the fact is very important
for potential software translators.

> In the same vein, did you consider using the ICU
> ResourceBundle and MessageFormat capabilities instead? If
> yes, what motivated your choice of gettext?

There is no problem to add support of any kind of 
dictionaries in future without breaking any API or ABI.

It is just a metter of writing code. In fact I would prefer
adding support of XLIFF in first place rather then support
ICU resource boundles.

Best,
  Artyom