Re: [tlug] Re: Piping stderr?

[Jiro SEKIBA <jir@example.com>]

At 28 Jun 2002 14:15:49 +0900,
Stephen J. Turnbull <stephen@example.com> wrote:
 
> Since there are (for practical purposes) no characters in other common
> character sets that aren't in Unicode, you must be talking about an
> uncommon character set.  That's not OK, because I can't use it.

 0xAAA1(GB18030).

> Unless you send it to me.  But if you do that, what difference does it
> make if you do it in the form of fonts + the native coded character
> set + CSI, or in the form of fonts + Cmap + Unicode private space?

 Plain text can't have fonts.

>>  ??????????  Sorry I don't understand your logic at all??
>> User can still use SJIS even removing SJIS locale by
>> filtering, which is YOUR WAY.  No?
> 
> If they have access to filters, yes.  How do you know they do?  Keitai
> denwa, for example.

 Since when do you start arguing general device...?
I thought we are talking based on PC(linux?) environment....
Can you show me the sentence you mentioned you start that kind of things?
And I'm still confused, filtering is your idea, not me...

 And about Keitai denwa, or let's embedded system whose computation
power is very low and memory is very limited, CSI is not suitable.
That kind a system must have target consumer and have to think about
which codeset is suitable.  In japan it was unfortunately SJIS ;-).
But is this what we are talking for long time?

> 1.  User is king; user does no work, user doesn't have to be logical,
>     user cannot necessarily use facilities on my host.
> 
> Therefore, I must provide facilities on my host (which typically is
> not the user's host) to handle all character sets known and unknown at
> the time of writing the program.

 It is CSI.  UTF-8 hard-coding program only support Unicode charset.
What are you talking about?

> 2.  I cannot assume that external encodings are "reasonable" (eg, "file
>     system safe" or "shell safe").
> 
> Therefore, I must translate external encodings to internal encodings.

 This is what CSI will do.

 - CSI       [API]  <CSI program>  [API]
    XXX encode ->  internal encode  ->  XXX encode

 - UTF-8 hard-cording
             <filter>   <hard-coding program>      <filter>
    XXX encode -> UTF-8 -> UTF-8(or UTF-16?) -> UTF-8 -> XXX encode

 Please study more about CSI first,  if you insists CSI is bad.
And I'm not insisting do not use UTF-8, also I am not arguing
which is the best codeset, O.K.?  I'm arguing program design, which is CSI.

-- 
Jiro SEKIBA | Web tools & AP Linux Competency Center, YSL, IBM Japan
            | email: jir@example.com, jir@example.com