Reset: Three Threads

Jecel Assumpcao Jr jecel at merlintec.com
Thu Nov 8 02:01:26 UTC 2001


On Wednesday 07 November 2001 16:52,  Justin Walsh wrote:
> [Dan Ingalls on Smalltalk's evolution]
> The above empirical scientific method is questionable.

Everything is questionable. My English, for example. But if you mean it 
doesn't work then I can't agree.

Are you aware of the differences between Smalltalk-72, -74, -76, -78 
and Smalltalk-80? And between traditional Smalltalk-80 and Squeak? The 
only other example I can think of for an improvement like that in a 
programming language is Forth, which isn't surprising since Chuck Moore 
is also willing to redesign his language once in a while.

Let me give you an example of Dan's method in action:

  1) make an observation - Randall Smith wrote a program called ARK 
(the Alternate Reality Kit) in Smalltalk-80 and tried to make 
everything as concrete as possible

  2) formulate a theory - ARK users didn't care too much about the 
object factories but liked to create new objects by cloning existing 
ones. Randall thought it might be possible to eliminate classes from 
Smalltalk.

 3) make a prediction that can be tested - Randall and Dave Ungar 
created the Self language (initially called Smalltalk-86) to see if the 
idea was practical.

Please describe how you would do better. I am not being provocative but 
really want to know.

> If you study the method closely is parallels the actual empirical
> method ....

I have no idea what you are saying here. I suppose you meant to write 
"it" instead of "is", but I still can't figure out what "method" and 
"actual method" refer to.

> Charles Richter says:
>
> http://www.ripose.com.au/ExtendedOsiModel.html
>
> What do you say?

He seems very confused about the relation between the OSI model and 
TCP/IP. The first pyramid diagram isn't correct. The OSI Application 
Layer is quite complex and divided into some confusing sublayers, so it 
is not surprising that he decided to ignore all that and add his own 
layers on top of the application. That doesn't make it right, however.

It isn't easy to see what this has to do with Dan's method, except that 
the ISO protocols designed using the "get it right the first time" 
method were a dismal failure while the Internet protocols designed 
empirically (see the IMP->TCP->web->XML evolution) eventually got good 
results.

-- Jecel




More information about the Squeak-dev mailing list