Re: Unicode status

Michael Bruck (
Sat, 30 Nov 1996 18:57:18 +0100

At 11:15 30.11.1996 -0600, Aaron Ucko wrote:
>[linux-console and linux-ggi added to Cc: list]
>Michael Bruck <> writes:
>> But as far as I remember Unicode is only present in keyboard handling code.
>> It's impossible (or at last not easy) to receive Unicode input in
>> applications.
>No, it's very easy; if your standard input is a VC, just
>#include <sys/kd.h>
>and call
>your program will then receive input in the form of UTF-8 encoded
>Unicode characters. This doesn't win you anything with the default
>keymap, though, and yields incorrect input when your keymap is for a
>character set other than ISO 8859-1.

Yes of course you can say the kernel not to care about input and
pipe characters (nearly) unprocessed into the application. But I
thought of a stable/robust solution. As far as I know switching
to K_UNICODE is risky. Nobody handles ^C, ^Z ... and if ^C would
be processed the user would end up with K_UNICODE in his shell. And
not everybody has a network connection to escape from this. (Yes
I know there is/are a/some trick(s) to get out of this.)

>> What I was thinking of was an extension to the output handling. Especially
>> it should be possible to write a graphics mode console (like the TGA
>> that is capable of displaying all characters. There are three + n problems
>> to be
>> solved to implement this:
>Although this can certainly be done, it is by no means clear that it
>should be; true text modes are, and will always be, faster than
>comparable graphics modes. Linux already allows 512-character fonts;
>while 512 is certainly less than 65536, it is plenty for many people.
>Nevertheless, it shouldn't be too hard to get GGI to support it if
>people are interested and don't mind the relatively slow performance.
>As for fonts, somebody has produced a free font with a lot of Unicode
>characters; I don't remember where it is, though.

What is GGI?

Of course graphics are faster, and I would still have a 386-25 with
Standard-VGA I wouldn't even think about wasting performance with this.
But in 1996 this of less concern. And who wants to run Linux as DNS-Server
on a 386-25 can use the standard textmode console.

What I mean is Unicode *support*.
Not "With Linux your *could* write an application that supports Unicode."
But "I can use Linux in a Chinese school to teach Russian and
allow them to exchange mail with a school there."
(An BTW Chinese Linux could be used by > 1GPeople
(where 1GPeople is 1024*1024*1024*People)).

About Win95, WinNT I can say:
"I just plug in my Latin/Russian/Arabic keyboard and start typing."
Of course there are not all programms Unicode-aware, but there are
enough that are. (f.e. Winword)"

At 14:19 30.11.1996 +0000, Alan Cox wrote:
>Its not hard, and the UTF-8 encoding used needs no extra magic to handle it.
>You can even use it in file names.

But 90% of all programs for Linux are displaying either a ? or \xxx instad.
They often don't even show characters that are in Latin1 encoding.