On Sun, Dec 14, 2003 at 10:48:57AM +0100, Christian Perrier wrote:
> First of all, sorry for just putting ideas in this mail and not
> propose any implementation. I'm afraid my skills are too low for
> proposing anything but really bad code. Not speaking of time
> lacking..:-)
> The current language list (which is the very first prompt a user
> "sees" when installing Debian) is not really satisfactory (imho, of
> course).
> The list is alphabetically sorted with all choices being translation
> of the english sentence "Choose this to proceed in xxxxxx"
> Thus, hungarian comes first (because of its leading "A"), then french
> variants (because I translated as "Choisissez" which comes before
> "Choose") and so on...
> There were a few proposals recently for changing this to simpler
> choices, for instance only language names.
> The variants also induce some confusion by macking the list a bit too
> long.
> Here are my proposal :
> -change this to a 2 stage process : first choose the language
> (e.g. "Français"). Then choose the country variant (is "Français" is
> chosen, the user chooses between Belgique, Canada, France, Luxembourg,
> Suisse)
I don't think a two-stage selection is all that necessary. That seems
like added complexity for little benefit; the list is only made a little
longer by including all locales in a single list, and the fewer
questions the better.
> -simplfy the sentence : just keep the language names (translated, just
> like they currently are)
Of course, the counterargument is that we can't expect the debconf
question itself to be usefully localized; so the user may understand
their language name, but without the complete sentence there may be
greater confusion as to the application of the language. Does it apply
to the installer only? To the default system locale once installed? To
the keyboard? Then again, I'm not sure the answers to these questions
are all that obvious even /with/ the current sentence. :)
> -find a rule for sorting languages :
> -English (which is the default) should come first
As long as this refers to the C locale (which would also be first if
sorted alphabetically by locale name -- using the C locale's sort rules,
that is :), that seems ok. If it's en_US or en_UK people are after,
those ought to be sorted like everything else. In which case, I think
the label for the C locale ought to be "international English", for
clarity.
> -Sort other languages. Below are some proposals:
> -alphabetically (not easy : where should double-byte languages go?)
> -by translation status in Debian:
> -first sorting key-->translation status for d-i
> -second sorting key-->combination of translation status for
> debconf templates and programs translations (by using
> http://www.debian.org/intl/l10n/ pages when they will be
> updated again
> -by numbers of people speaking these languages in the world. Many
> references may be found on this topic. For instance, by googling
> a bit, I found http://www.photius.com/rankings/languages2.html:
> -Chinese, Mandarin
> -Spanish
> -English
> -Arabic
> -Bengali
> -Hindi
> -Portuguese
> -Russian
> -Japanese
> -German
> -Chinese, Wu
> -Javanese
> -Korean
> -French
> -Turkish
> -Vietnamese
> .../... we have 50 languages listed
> This reference is an ethnological reference which lists the
> number of people *really* speaking the given language. This
> is not a sum of the populations of countries for which
> the given language is among the official languages of the
> country (I found that statistic also, but only for the Top10
> languages : http://www.infoplease.com/ipa/A0775272.html#A0774735)
> Does somebody have comments on this?
I believe that the most useful sort order for a user *trying to find his
language in the list* is alphabetically by locale code, with locale
codes listed explicitly in front of the translation. Most ISO locale
codes are derived from the language's own name for itself, so assuming
at least basic familiarity with Latin sort order, this would seem to be
the biggest UI win for the most users.
CJK is an exception; not only are speakers of CJK languages least likely
to be familiar with Latin sort order, their language codes are less
likely to correspond to the native name for the language. ('ja' for
Japanese comes from the Western name for the language, as does 'ko' for
Korean, IIRC; and I'm not sure of the derivation of 'zh' for Chinese.)
If someone can think of a sort algorithm that would better serve CJK
users without compromising the utility to the users of the much more
numerous (and harder to identify at a glance) Latin locales, I'd be
interested to hear it.
Sorting by translation status in Debian is only useful to the user if he
*knows* that's the reason for the sort order. Otherwise, he's just
going to get frustrated trying to find his preferred language, and then
get frustrated some more when half of the installer runs in English.
Sorting by number of speakers world-wide is interesting, but again,
would cause frustration for anyone not listed on the first page. Also,
several of the top world languages currently have little or no l10n
support in d-i, and the number of *Debian-using* speakers of those
languages is not necessarily proportional to the number of speakers
world-wide.
One possible hybrid sort would be to choose the first five or six
"preferred" languages (by whichever criteria) to list first, and then
sort the remainder alphabetically by country code -- this lets most of
our users pick from the top, and leaves enough space on the screen for
other users to observe the sort order.
--
Steve Langasek
postmodern programmer
Attachment:
signature.asc
Description: Digital signature