Path: utzoo!attcan!uunet!dino!ux1.cso.uiuc.edu!ux1.cso.uiuc.edu!m.cs.uiuc.edu!johnson
From: johnson@m.cs.uiuc.edu
Newsgroups: comp.lang.smalltalk
Subject: Re: PPS R4 under OpenWindows?
Message-ID: <5600011@m.cs.uiuc.edu>
Date: 27 Sep 90 12:02:00 GMT
References: <174602@<1990Sep17>
Lines: 69
Nf-ID: #R:<1990Sep17:174602:m.cs.uiuc.edu:5600011:000:3980
Nf-From: m.cs.uiuc.edu!johnson    Sep 27 07:02:00 1990


I am going to speculate about ParcPlace's strategy with regard to their
newest version of Smalltalk.  I don't have any inside information (well,
not very much!) but this sounds logical.

First, it has been common knowledge for many years that MVC has needed 
rewriting.  The reason (I heard) that the fourth book of the Smalltalk
series, the one on building user interfaces, never came out was because
the Xerox people had decided that they didn't want to document the
wrong way, but wanted instead to fix it.  This was around 1983.
I can remember an early OOPSLA (was it 1987?) where Peter Deutsch talked 
about his latest rewrite of MVC and why it wasn't a big enough improvement
over MVC to go through the trauma of changing.  ParcPlace knows how much
of a problem it is to users when they change standards, but they also
know how much better off everyone is with a good standard instead of a
bad standard.  MVC was the very first user interface framework, but the
last 10 years have seen a lot of research on the topic, and there have
been many new user interface frameworks, such as MacApp, the Andrew ToolKit,
and Interviews.  Each of them borrows a lot from MVC and also improves
upon it.  MVC is showing its age, and has needed to be redone.

For example, MVC is the only u.i. framework that I know about that uses
polling instead of being event driven.  This causes major problems if
you try to do background processing.  When I look at MVC as a research
project, I would say that one of its main achievements is to prove
conclusively that polling is the wrong way to build a u.i. framework!

In addition to fundamental design flaws, some of the code in MVC is
of very poor quality.  In particular, classes ParagraphEditor and
Paragraph are a mess.  One of my students rewrote them a couple of
years ago, so I know that it is possible to make them much easier
to subclass and customize.

I could give a long list of things that are wrong with MVC.  However,
I think the thing that is causing ParcPlace to finally rewrite it is
that it doesn't work well with foreign window systems, but reserves all
the responsibilities and priviledges of a window system for itself.
My guess is that it will finally be able to work with X over a network.

I'm sure that ParcPlace has tried to make the new MVC compatible with
the old one.  They might have done this with a compatibility package,
or maybe they just ensured that the new system is similar to the old.
In any case, their customers are always complaining about the differences
from one image to the next (I heard a number of late hour gripe sessions
at OOPSLA'89 whenever a ParcPlace senior designer would wander by--
they defended themselves well).  The problem is that compatibility is
opposed to fixing design flaws.  Bertram Meyer described the problem
well in his article in the current CACM.

It is clear to me that the reason that ParcPlace is pricing the new
version of Smalltalk so high is that they don't want people to buy it.
They will continue to sell the old version (2.5) until they get the
bugs worked out of 4.0.  By "bugs", I don't mean broken code as much
as unnecessary incompatibilities with the old MVC and the like.  They
probably intend that only a few people will buy 4.0 at first; these
will be the people who need it to start developing products using it.
In a year or so they will reduce the price to where 2.5 is now and
deemphasize 2.5.  As I said before, this is all speculation, but it
certainly seems logical to me.

If I were a commercial developer of Smalltalk applications then I
would call up ParcPlace and ask them what the price of 4.0 will be
a year from now before I made any hasty decisions about dropping
Smalltalk-80 because of its price.  I can't believe that they think
that they can sell Smalltalk-80 at this price, so I think that it
is a marketing strategy of some sort, and that the price will eventually
go down.

Ralph Johnson -- University of Illinois at Urbana-Champaign