Skip site navigation (1)Skip section navigation (2)
Date:      Thu, 11 Apr 1996 14:52:00 -0700 (MST)
From:      Terry Lambert <terry@lambert.org>
To:        alk@Think.COM (Tony Kimball)
Cc:        terry@lambert.org, hackers@freefall.freebsd.org
Subject:   Re: Lesstif (motif compatible) package.
Message-ID:  <199604112152.OAA05099@phaeton.artisoft.com>
In-Reply-To: <199604112126.QAA01909@compound.think.com> from "Tony Kimball" at Apr 11, 96 04:26:31 pm

next in thread | previous in thread | raw e-mail | index | archive | help
>    So will X-builder turn out Tk code?  It turns out Motif code...
> 
> xf turned out tcl7.3/tk3.6 code.  I understand there are patches to
> support a more modern incarnation as well, but my sense is that people
> doing tcl/tk find gui-builders largely superfluous, mostly because the
> geometry managers handle the layout issues that people use gui
> builders to take care of.  xf would need substantial enhancement in
> order to parallel any of the commerical resource editors.

This was my point.  The level of quality is just not there for the
non-commercial products (if it were there, their authors would be
sill to have expended such an outrageuos amount of effort such as
these things require, without taking them commercial -- that's
where new commercial products come from).


> I know you are quite fond of GUI builders.  My experience is perhaps
> less glowing than your own.  (Mine is based on about a year spent
> inside Visual C++, and a few months inside AppBuilder and TCL on the
> Mac, doing all GUI-oriented code.)  In my mind, any one specific
> commercial GUI builder does not suffice to make an argument in favor
> of technical accomplishment, because it is ghetto technology.

GUI builders allow incompetent people to build user interfaces, and
can, to some small extent, enable competent people to be more
productive.  In this respect, they are much like prototypes and other
ANSI C features that would have been better placed in the linkers
on the theory that since compiler users outnumber compiler writers
100:1, anything that saves the average compiler user 1 hour is worth
100 hours of compiler writers time to make work.

I disagree with the current cognitive models in use in most recent
GUI-builder products because of the resulting code not being well
abstractedat the right layers for code portability in favor of
making it easier for the GUI-builder-writer.

One could easily make the case that a GUI-builder decription (GUI
schema) should be portable across environments with no changes.

Use of Microsoft foundation classes in the resulting code, and
the programmer being encouraged to throw the code all in one
directory and one file, and the internationalization techniques
commonly use by graphics programmers all conspire against code
portability.

Visual BASIC is the absolute worst offender.  OCX's encourage minor
variations in the style from one application to another which makes
user training unportable.


All this said, that doesn't mean that GUI builders are not a good
idea, even if the examples I can show you are all either implemented
by morons with no concept of cognitive psychology and other human
machine interface issues (proprioception in the mouse/pointer
linkage, etc.), or priced beyond the level of general access.


> X-builder can be used by perhaps 5% of all X developers.  X developers
> are perhaps 5% of all GUI developers.  That's pretty irrelevant
> to the bigger picture.  Likewise xf is irrelevant, but at least it
> could potentially be made relevant because it is portable to non-X
> environments and because it is free, whereas X-builder could not.
> Besides which, these fancy resource-editors just don't account for
> much, over the life-cycle of any real application.  (They're great
> for little one-offs, though.)

Yes.  This is really the problem with the architectural model they
imply must be used in applications, which is a problem with the
builder implementation, not with the idea of builders.


>    Except that Motif drag-and-drop interoperability is part of the X/Open
>    Common UNIX Standard compliance requirements.
> 
> A de jure standard which is closed in practice is just a 
> marketing tool, not a standard.  Microsoft has played that game
> for years.  OSF is doing it too.

Yes.  This is my problem with Spec 1170.  This is my problem with ABI
and POSIX compliance certification being so expensive.  It is a club
membership buyin.

I would argue, though, that Motif is more de facto than de jure; it's
popularity is one of user demand (evidence: Sun's switch to Motif
from OpenWindows).

You could make the same argument for Microsoft's Win32 being a standard
by conspiratorial fiat -- but it is no less a standard for it.


> A de facto standard which is open in practice is to be preferred,
> on technical, moral, economic, and aesthetic grounds alike.

Agreed.  This is why it is painfaul to see the Lesstif people cutting
corners that put them in legal jeapordy, and using an unmodified
LGPL, putting their users in legal jeapordy (unless we all go to ELF
and modify further the Linux ELF shared library implementation, etc.).
All a lot of pain to resolve a problem that shouldn't have come into
existance in the first place.


>    > The small fraction running Motif.  It's too big to ship static
>    > executables.  
> 
>    Require shared libraries.
> 
> Precisely my point.  Teeny weeny tiny market slice.  Not realistic for
> commerical products, except in niches.  Admittedly, those niches can
> be quite lucrative, but they are still little niches.

You don't argue that X is teeny (relative to the UNIX market -- the
mismanagement of the UNIX market by the big 3 is another discussion
entirely).

A common GUI is only possible if everyone adopts the same standard,
and the big 3 aren't about to dump Motif and adopt Tk.

> Hey, I'd never cripple myself with a straightjacket like that.  I'm
> making a real-world argument from personal goals and values.  
> That being the case, this is no longer appropriate to
> "hackers".  Thus I shall refrain from further follow-ups on this
> subject.

8-).


> You can have the last word.

8-).


					Terry Lambert
					terry@lambert.org
---
Any opinions in this posting are my own and not those of my present
or previous employers.



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199604112152.OAA05099>