Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!wuarchive!emory!gatech!bloom-beacon!dont-send-mail-to-path-lines
From: matt@saber.COM
Newsgroups: comp.windows.x
Subject: Re:  Question for net.views column in UNIX Today!
Message-ID: <9103300619.AA12443@rapier>
Date: 30 Mar 91 06:19:44 GMT
Sender: tytso@athena.mit.edu (Theodore Y. Ts'o)
Organization: The Internet
Lines: 94

> 	*******************
> 	*   QUESTION #2	  *
> 	*******************
> 
> 	Is a single GUI standard really necessary?

Depends what you mean by that.

Is a single standard for application appearance and behavior necessary?

Probably not.  For one thing, none of the existing standards is really
adequate for all classes of applications one can envision, or even for 
all existing applications.  For another, personal preference will always
play an important role -- witness the radically different opinions you
find in the field about Open Look and Motif, even among people who 
have used both extensively.  

User interfaces are a bit like architecture or interior design -- there 
are some fundamental principles one should follow (but which are often
all but ignored in software design), but there's also a lot of style, 
taste, and religion.  

It is not important that all applications everywhere look and feel the 
same -- the only thing that is important is that all of the applications 
used by a *single person* look and feel the same; moreover, the look and 
feel should be whichever one *that individual* prefers.

Trying to dictate a single standard for application behavior is probably
both unnecessary and counterproductive; even were it not, it seems clear
that none of the existing contenders (Open Look, Motif, Windows/PM) comes
anywhere near what would be required in terms of either expressive power 
or useability from a human factors standpoint.

What is necessary is an attempt to bring some order out of the chaos that 
currently surrounds user interface toolkits and application programming 
interfaces.  

The investment required to develop a user interface to match any particular 
GUI specification is quite high, and it *does not decrease* when you go 
to apply a second GUI "standard" to the same software.  This has a number 
of pernicious effects on the industry.  

One is that application developers are often forced into choosing sides 
in the GUI wars, which are based more in politics than in either technial
or human factors superiority.  

Another is that developers who must support multiple GUI standards generally 
do it by reimplementing their interfaces for each standard.  This wastes 
technical and engineering resources that could be better used in designing 
*better* user interfaces, instead of building *different* user interfaces.

Finally, the fact that every behavior specification has one (or more)
different API's discourages advancement in the state of the art in user
interfaces.  Because most of the existing toolkits are tied so closely
to a particular GUI specification, developers who must worry about real
deadlines and market windows are discouraged from experimenting with 
fundamentally different application styles.  Standardization of the GUI
in its current form leads to stagnation.

Toolkits that support multiple application styles with a single API (or
better still, that supported runtime selection of the application style
by a single executable image), would go a long way toward encouraging
better use of resources, promoting experimentation and advancement in
user interface styles, and allowing users some degree of *choice* about
the style of application behavior they prefer.

There are those who will say this is too hard, but they are wrong.  Real 
progress is being made in applying a single API to the Open Look and Motif 
specifications as well as others, and we are beginning to understand what 
it means to "be" a user interface in a larger sense, and how to abstract 
the higher level components of an interface in a way that lets us deal 
with them independent of a particular visual representation or interaction 
model.

Rather than leaping headlong onto the bandwagon of premature standardization
of the "look and feel", we should be searching for ways to provide uniform
*programming interfaces*, so that application developers need not choose
only *one* "look and feel", but are free to develop new and *better* kinds
of computer/human interaction without being forced to throw away all of 
their prior investment.

	- Matt Landau
	  User Interface Engineering Group
	  Saber Software Inc.

	  matt@saber.com

	  (617) 876-7636

	  Disclaimer:  There opinions expressed herein are the opinions
		       of the author, who is solely responsible for their
		       content and expression.  These are not necessarily
		       the opinions of Saber Software Inc, or any of its
		       officers, other employees, or agents.