Re: man-db locale support
- To: Michael Sobolev <mss@transas.com>
- Cc: debian-i18n@lists.debian.org
- Subject: Re: man-db locale support
- From: Fabrizio Polacco <fab@prosa.it>
- Date: Fri, 1 Oct 1999 12:07:24 +0300
- Message-id: <[🔎] 19991001120724.N8737@none>
- In-reply-to: <19990930204201.B27277@transas.com>
- References: <Pine.GSO.4.10.9909231516040.29462-100000@info> <y5aso44qwm9.fsf@xlj06203.nifty.ne.jp> <y5aaeq65veh.fsf_-_@xlj06203.nifty.ne.jp> <19990929203801.G8737@none> <19990930035421.A15638@transas.com> <19990930170129.K8737@none> <19990930175513.A25266@transas.com> <19990930181437.L8737@none> <19990930204201.B27277@transas.com>
On Thu, Sep 30, 1999 at 08:42:01PM +0400, Michael Sobolev wrote:
> >
> Well, if look on the code of man program, we will see that for every directory (path)
> in MANPATH it adds path/lang *before* this path. So, from MANPATH=/usr/share/man,
> we'll get MANPATH=/usr/share/man/ru:/usr/share/man, which means that if we do not
> have a translated manual page the default one will be shown. The patch I sent
> you does the same, but for LANG=country_TERRITORY.ENCODING,VERSION, it adds all of
> them from the most specific to the least specific. So everything should continue
> to work. No?
Humm, nice patch, but ... it brakes japanese :-(
Maybe you've missed my last 69k (now 69m) which had something similar:
set the LANG to ja_JP.ujis and it will search .../man/ja_JP.ujis/manX,
.../man/ja_JP/manX, .../man/ja/manX which is what we wanted, ... or
not?
What should be the correct behaviour?
giving a locale like "ja" serch for all the pages in any man/ja*/manX
place (and giving a locale like "ja_JP.ujis" should it find a manpage in
man/ja/manX or default to english?).
In theory this looks like the most reasonable.
But people asked for the contrary (if I didn't misunderstood): giving
"ja_JP.ujis" as LANG, display pages in man/ja/manX (actual behaviour).
> What I mean is that if one specify -t latin1, one gets a decent result, nothing
> more. There is a problem with -t latin1, as the `?' character appears as
> something unexpected. But to my knowledge, nobody hacked groff for supporting
> -t koi8-r (or anything similar).
This is another big source of trouble: it looks like some locales gets
good results using some setting _because_ they use some other tool
hacked to do this.
It's like fixing a bug introducing another bug in a different place
which tries to put thing back in place.
fab, wondering to ask: "is sgml (docbook) able to render your
language/charset on your actual linux screen?"
--
| fab@pukki.ntc.nokia.com fpolacco@prosa.it fpolacco@debian.org
| 6F7267F5 fingerprint 57 16 C4 ED C9 86 40 7B 1A 69 A1 66 EC FB D2 5E
| fabrizio.polacco@nokia.com gsm: +358 (0)40 707 2468
Reply to: