Bits about countrychooser and languagechooser
In this rather long mail, I'll try to summarize the ideas which float
around about these topics : choosing language, country,locale and all
this stuff. Things are currently not clear for various reasons
(including some by myself..:-)))
I had some discussion this "morning" (UTC) with pere about this
thing. Really interesting and constructive, indeed.
The original idea which lead to countrychooser was mostly brought by
critics I read and heard about the original language selection system:
-a bit "messy" (no real sorting)
-why some language/country combinations and why not some others ?
-which country to show for widely spoken languages (English, French,
Arabic) ?
This is why I first came up with a "languagechooser_ng" (CVS branch of
languagechooser) and a new countrychooser package.
The first allows selecting a language.
The second allows choosing a country (or region, Carlos...:-)))
pere (Petter Reinholdtsen for those not fond of nicknames) is not
really enthusiast about this for the following reasons (pere, correct
me if I'm wrong):
-one more question asked to users
-confusion about country : which country ? The one you live in, the
one to use for choowing keyboard and mirror ?
-confusion about locale settings resulting from this choice.
Petter is (imho) indeed right for some points. I think the
additionnal question issue shouldn't be a blocking point. We probably
can save one question elsewhere, for instance by lowering the mirror
country question.
He already answered the sorting issue by commiting a nice change to
languagechooser this morning. Just drop an eye to it.
The locale problem is more problematic. One of my ideas while
proposing languagechooser/countrychooser separation was allowing the
user to choose among all possible locales (a locale is a combination
of a language, a country and a charset.....the country being the one
you live in.....at at least the one you use the "regional settings"
such as currency unit, decimal separator, etc.).
The current countrychooser is not satisfactory when it comes to
this. I've made some attempts for getting a correct locale after it
runs...but no locale validation is made.
There are 239 "valid" (supported) locales in Debian (from
/usr/share/i18n/SUPPORTED) and this is likely to change often.
So, I think that having all these listed in languagechooser could
make it pretty unusable. Look at all SUPPORTED es_* locales, for
instance. Or imagine the future fr_* list when all african locales
will be supported.
Language which don't have a clear "major" country, such as Arabic,
also explain this concern.
This is why I'm still thinking countrychooser has an interest, mostly
for choosing the "country" part of the locale, but also for giving a
good default for keyboard and mirror.
Otherwise, we would have to extend the languagelist over and over,
making it slowly unreadable.
But, maybe, simplifying languagechooser was not the best idea : we
could keep the current list, with the most common language_COUNTRY
combinations.
But, in the same time, add an "Other" entry for all languages. This
"Other" entry would then trigger countrychooser (maybe showing a
subset of the whole country list....with all *possible* countries,
according to /usr/share/i18n/SUPPORTED).
Something like this:
ar - (???? ??? ????? ????? ??????? (?????? ??????
bg_BG - Изберете тази опция\, за да продължите на български
.../...
en_AU - Choose this to proceed in English (Australia)
en_GB - Choose this to proceed in English (United Kingdom)
en_US - Choose this to proceed in English (USA)
en - Choose this to proceed in English (Other countries)
.../...
fr_BE - Choisissez ceci pour continuer en français (Belgique)
fr_CA - Choisissez ceci pour continuer en français (Canada)
fr_CH - Choisissez ceci pour continuer en français (Suisse)
fr_FR - Choisissez ceci pour continuer en français (France)
fr - Choisissez ceci pour continuer en français (Autres pays)
.../...
All languages without XX would then trigger countrychooser for
choosing the correct country (ideally by grepping
/usr/share/i18n/SUPPORTED) for possible other countries. For instance,
for English:
bubulle@mykerinos:~> grep -e "^en_" /usr/share/i18n/SUPPORTED | cut -b 4-5|sort|uniq
AU
BW
CA
DK
GB
HK
IE
IN
NZ
PH
SG
US
ZA
ZW
countrychooser then builds a "on the fly" template and shows all these
choices...in English of course, and with country names.
With such system, we have only one question for most common
combinations and we keep all possible locale settings.
Folks, please react to this. I know this is not a highly techical
issue, but this is probably an important issue anyway....
Reply to: