Re: gjiten 0.8

["Stephen J. Turnbull" <turnbull@example.com>]

>>>>> "Jim" == Jim Breen <jwb@example.com> writes:

    Jim> But up to a point, doing silently what the user means
    Jim> (wants?) is the way we should be going.

Have you been reading Asimov "telepathic robot" stories again?  "Doing
what the user means/wants" can only be implemented as "Doing what I in
my God-like wisdom think the user _should_ mean/want."  Typically even
that understanding is buggily half-implemented.

If the user doesn't like warnings, but prefers to have his software
surprise him, I have no objection to implementing a
"PLEASE_SURPRISE_ME" variable, option, or whatever.

But personally, I like warnings (2>/dev/null works fine when I don't),
and I have "POSIX_ME_HARDER" set to impose some order on those helpful
chaps in Cambridge MA.

    Jim> HAving software that says "Naughty, naughty. You can't
    Jim> display Japanese here. There is a variable deep in the system
    Jim> which declares you to be irrevocably speaking English with a
    Jim> Merkin accent, so I am duty bound to stop you doing anything
    Jim> else" might make some people feel better but it leaves me
    Jim> cold.

Well, yes, that's how I'd write it.  I'd probably ask Wiley to review
it and catch any latent softheadedness left, too.

But didn't I make it plain that as far as I'm concerned _your_ program
can do whatever it wants, as long as it tells the users with more
brains than money that it's doing those things on purpose, and,
preferably, how to achieve more flexible behavior?

I really fail to see why having a program that is 95% right but that
5% is inherently unfixable because it's hard-coded is a good idea.

Sure, a program that can "guess" what a clearly clueless luser wants
is useful to far more people.  But the reason I don't use MS Windows
(built my first XEmacs on it last night, by the way, YOW!  We ARE
COOKIN' with AXLE grease now!) is neither religious opposition to
closed source nor (my personal religion) opposition to monopolies
restricting supply.  It's the second-guessing nature of Windows
software, in places I do _not_ want to be second guessed.  And the
useless nature of the Help system, which tells you to ask the system
admin any hard questions.  Hello?  This _is_ the sysadmin speaking!

I just don't see how providing the warnings for those who can make
good use of them really impairs the experience for those who can't.

    Jim> here was some tut-tutting recently about kterm ignoring
    Jim> locales. Good on it!

It used to hose your Big5 and KOI8-R displays if you were not careful,
which it _can_ do rigth if you know how to force it.  It's documented
as being capable of doing multilingual stuff, too, so you can't tell
me that "that's not the responsibility of a Japanese program."

    >>> Tubers have nothing to do with it.  Any POSIX system
    >>> implicitly has all LC_* categories set to "C" by default.

    Jim> And tell me again how the I18N locale system isn't
    Jim> chronically broken?

Huh?  Since when is having an absolutely innocuous default "broken"?

The POSIX locale system is "broken" mostly because all the
implementers looked at it and said "I can do better than the
standard."  And they were right -- except that they all chose to do it
differently, non-portably, and usually inappropriately for almost all
locales (in the broad sense) other than their own.  Leaving us with a
system that really sucks.

Sure, POSIX has lots of problems for multilingual applications and the
like.  But that's not the kind of user we know how to second-guess
yet.  And they screwed up standardizing on catgets (OK, there's one of
the wrappers I was talking about, gettext) and some of the internal
functions (mostly much less important than gettext, though, see the
glibc docs on nlsinfo functions, IIRC).  The Unix 9x and C 9x
standards fix a lot of those problems, too.  Unfortunately, the
standards are not cheap, let alone free (in either sense).  (That's
really the most broken thing about POSIX, as compared to say Unicode:
you can't get the standard for US$49.95 from Amazon.)

-- 
University of Tsukuba                Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Institute of Policy and Planning Sciences       Tel/fax: +81 (298) 53-5091
_________________  _________________  _________________  _________________
What are those straight lines for?  "XEmacs rules."