[Pkg-crosswire-devel] ICU use and debian/tmp/usr/lib/sword/1.5.11_icu_3.8.1/translit_*.res files

[Matthew Talbert]

> Matthew Talbert wrote:
>
>> This library adds, besides the transliteration, character ordering
>> according to specific locales. This is important for creating search
>> indexes, creating dictionaries, etc.
>
> Hmm, isn't character ordering per locale done in Linux/Unix/POSIX by
> setlocale(3) and then using the strcoll(3) function?  I'm probably just
> being naive or out of date... I've not done any serious C programming
> for some years.
>

Yes, it is. The existence of ICU (and a few other similar projects) is
because the POSIX standards do not cover the entire range of locales
with their unique peculiarities. ICU uses the data from
http://www.unicode.org/cldr/ for this which supersedes what is used by
glib (for example) and also differers from it in some respects. To be
honest, I can't tell you what exactly the differences are.

Apparently front-end developers in the past thought that their
platforms provided sufficient tools for this, and there certainly have
been the stability issues I mentioned with ICU. As SWORD itself moves
to doing some fancier things with ICU and increases the stability,
this situation will probably change.

So, to recap, I've never actually heard of any (current) problems that
specifically needed ICU to solve as opposed to the standard POSIX
tools either for module developers or for the front-ends. And again,
GnomeSword users won't see *any* difference with a libsword compiled
against ICU except for the potential stability issues.

Perhaps Peter can comment on how exactly ICU provides a benefit over
the built-in locale ordering and collating provided by POSIX for
module developers.

Matthew