Xref: utzoo comp.edu:2624 comp.software-eng:2344
Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!cs.utexas.edu!uunet!portal!cup.portal.com!Jerome_V_Vollborn
From: Jerome_V_Vollborn@cup.portal.com
Newsgroups: comp.edu,comp.software-eng
Subject: Re: CS education
Message-ID: <23988@cup.portal.com>
Date: 13 Nov 89 04:49:10 GMT
References: <11064@cbnews.ATT.COM> <6961@hubcap.clemson.edu>
  <16028@duke.cs.duke.edu> <23871@cup.portal.com> <20410@brunix.UUCP>
Organization: The Portal System (TM)
Lines: 33


Scott Meyers (sdm@cs.brown.edu) writes:
>
>Just out of curiosity, given that he was doing research and wasn't 
>necessarily sure what the final specification of his system was 
>supposed to be, what design method would you have recommended he
>employ?
in response to my posting about a PhD candidate doing research 
in machine control (the PhD was in computer science from University 
of California at Berkley).

Personally I have had good results with object oriented design ala
Grady Booch as extended by EVB Software Engineering (contact Ed 
Berard at Berard Software Engineering or Brad Balfour at EVB Software 
Engineering for details).  There are two main points in response to 
Scott's questions:  (1) the object of the research was machine control 
(which was well done in this case) not building software; and (2) time 
spent debugging software was time stolen from machine control research.  
The person has developed a knowledge base that would be valuable to 
my company but I find the attitude that if you don't know everything 
about a problem you can't use an explicit design method to be almost 
criminal (product liability juries find it negligent).  In virtually 
every project I have done, I did not have all of the information 
needed.  I have had to go ask questions, make judgements about what 
I thought the users wanted, and had to go back and modify the software 
to do what the users really wanted (but couldn't coherently state).
However for the last ten years I have used explicit design methods 
starting with Yourdon and Constantine's method (aka SA/SD).  I am 
still studying formal methods but haven't applied one yet.

				Jerome Vollborn
				(Jerome_V_Vollborn@cup.portal.com or
				 uunet!lci386!jerome)