RE: tlug: Fw: Could Linux Kill NT?

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

>>>>> "Matt" == Matt Gushee <matt@example.com> writes:

    Matt> Indeed. The GUI is going to be a tough one, though. There
    Matt> was an interesting discussion on slashdot.org about the
    Matt> standardization issue, with Jamie Zawinski saying that GTK
    Matt> is a waste of good programming talent, and that LessTif is
    Matt> the way to go, being standard on commercial UNIX.

    Matt> I have to wonder, though. JWZ says that commercial UNIX
    Matt> vendors will never leave Motif unless they find something
    Matt> obviously, hugely, better (whereas GTK is just a little
    Matt> better) and that we ought not to turn our backs on our
    Matt> "natural allies" (not sure if that's a quote or a
    Matt> paraphrase) at Sun, SGI, etc. ... all of which makes a good
    Matt> deal of sense.

    Matt> But with all due respect to one of the movers and shakers, I
    Matt> think he's wrong. Maybe it depends whether the
    Matt> <CLICHE>paradigm shift</CLICHE> is a top-down (i.e., driven
    Matt> by IT managers and so on) or a bottom-up affair. People who
    Matt> run networks may choose Linux for performance and stability,
    Matt> but for most individual users, even if they appreciate those
    Matt> benefits, they're not going to migrate en masse to Linux
    Matt> unless it has a great GUI. And regardless of how developers
    Matt> view the relative bugginess/stupidity of the various
    Matt> toolkits, from an end user's standpoint (at least from mine)
    Matt> Motif doesn't cut it, not by a long shot.

I think you're missing Jamie Z.'s point.  Windose 3.1 is very nearly
Motif Style Guide-compliant (see OReilly vol 5B or whatever); I can't
speak for Windose 95.  The UI stuff on Unix is not a problem with
libmotif.so (or whatever it's called); it's a problem with developers
who use Motif widgets in non-Stype Guide-compliant ways and sys admins
who f**k with app-defaults.

Have you used a proprietary Motif supported by a money-hungry vendor?
Sun's worked (when we had Sun) most of the time, and when it didn't,
it was usually a local config error.  Of course, I lived in Emacs, so
about all I did with Motif was to bang on the Emacs icon ;-)

    Matt> Windows is a set of 32-bit extensions for a 16-bit shell for
    Matt> an 8-bit ... yeah, we know all that. But give credit where
    Matt> credit is due. When you click on a Windows icon, it jumps up
    Matt> and does something (even if what it does is stupid); when

Bites me in the ass usually.

It is possible to misprogram Windose apps.  The canonical Lose32
example app (Freesilt) has a really annoying set of stupid modal
dialogs.  You _must_ answer yes/no to the stupid end-of-game dialog
("Do you want to play again"), there isn't a quit option.  When using
the "click-on-source, click-on-target" interface, if you miss the
target (either because of slipping on the mouse button or a think-o or
colorblindness) you lose the selection and have to start over again.
(This is especially annoying on a touchpad.)

You can't solve these problems with a new toolkit.

Note that one of the few pieces of Free (I think it is, although I've
never actually seen the source code) to come out of M$ violates user
interface principles all over the place.  Are you surprised?[1]

    Matt> you hit the tab key, the cursor usually moves in a
    Matt> reasonable way ... all sorts of normal and intuitive GUI
    Matt> behavior that is annoyingly broken in Motif. And the stupid
    Matt> backspace/delete thing ... alright, I guess it's not Motif's
    Matt> fault, and it's a trivial fix for most of us. But if we want
    Matt> mass appeal, those sorts of problems have to go away.

If it's a trivial fix for _us_, then Scott Stone and Ian Jackson and
watzisface at cdrom.com can do it too, in app-defaults or using the
Motif keyboard manager, and we'll never see it again.

Most of these problems do NOT require a new widget set to fix.  In
fact, many of them will need to be fixed AGAIN in new and incompatible 
ways (The Gospel according to Murphy, Chapter 1, Verse 1).

Question: Is there a GTK Style Guide?  If not, you can bet that the
suite of ported-to-GTK apps will flunk the consistency test as badly
as Lesstif does.

    Matt> Manuel Chakravarty writes:

    >> I am far from being a Motif fan, so I understand your last
    >> remark, but I am unsure about GTK.  What's the problem with it?
    >> And if it is not much better than Motif now, doesn't it have
    >> the possibility to develop -- in contrast to Motif?

It has the possibility to rapidly evolve, it's true.  However,
somebody can (eventually) make what we want of Lesstif, although it
will be a massive pain in the butt doing that AND maintaining
backwards compatibility with the Motif standard, preferably in a way
that is transparent to the developer (a la Xaw3d).

This is more constraining, and you don't get to fix the design bugs.
But your library will be used by a lot more developers, and thus users.

    Matt> Hmm, you may be right. The GTK thing ... I was just sort of
    Matt> passing along JWZ's remark without too much thought. His
    Matt> point was that, even if it is better than Motif, it would
    Matt> have to be a spectacular improvement for the commercial UNIX
    Matt> vendors to even think about changing. Knowing very little
    Matt> about the commercial UNIX world, I just have to take his
    Matt> word for that.

There is an immense amount invested in Motif; it's not quite as bad as 
y2k and Cobol, but not more than a couple of orders of magnitude off.

    Matt> Personally, I think GTK is great compared to other *NIX
    Matt> toolkits. However ... and maybe this is just me, but I'm a
    Matt> keyboard-oriented guy ... so I think that a good GUI is one
    Matt> where you have a choice of using the keyboard or the mouse

A concept invented and implemented by ... OSF/Motif.  Hmmm....

Of course, you do need to go to the programming effort to bind the
hotkeys in Motif, too.

    Matt> for *everything* (except, of course for operations like
    Matt> drawing/painting images). So while I think dialog boxes are

Painting by keyboard was supported by DESQview/X.

    Matt> a good visual aid, you should be able to operate them
    Matt> without ever touching the mouse. GTK doesn't do that
    Matt> ... though I suppose it wouldn't be too hard to make it so,
    Matt> if one wished.

Why bother?  Motif already does that.

But the developer has to add a few lines of code to define some `tab
groups' and the like.  The keyboard and the normal translations are
all handled automatically by the library, though.

    Matt> But I guess the main point was that 'good' and 'viable' are
    Matt> -- as we all know -- not the same thing. Is GTK viable in
    Matt> the marketplace? Your guess is at least as good as mine.

Not for years.  If you define "marketplace" the way Jamie Z. does.  It's
probably viable now in Linux only.


Footnotes: 
[1]  Unlike my usual anti-M$ rant, I'm implying that it's the
free-ness of Freecell that is the problem, not the M$ origin.

-- 
University of Tsukuba                Tennodai 1-1-1 Tsukuba 305-8573 JAPAN
Institute of Policy and Planning Sciences        Tel/fax: +1 (298) 53-5091
--------------------------------------------------------------
Next Nomikai: 17 July, 19:30 Tengu TokyoEkiMae 03-3275-3691
Next Meeting: 8 August, Tokyo Station Yaesu central gate 12:30
featuring Linux on multiple platforms:
i386, Sparc, PA-Risc, Amiga, SGI, Alpha, PalmPilot, ...
--------------------------------------------------------------
Sponsor: PHT, makers of TurboLinux http://www.pht.co.jp