🚀 go-pugleaf

In article <3298@thorin.cs.unc.edu>, omondi@unc.cs.unc.edu (Amos Omondi) writes:
< In article <Jun.30.04.56.23.1988.24272@aramis.rutgers.edu<, webber@aramis.rutgers.edu (Bob Webber) writes:
< < 
< < Hmmm.  I am not sure quite what you mean to be implying there.  The
< < modifications were quite ``real.''  For example, there was a report
< < authored by ``Anonymous'' entitled ``Description and Use of the ENIAC
< < Converter Code'' reviewed in 1950 (vol 4, pp. 150-151 of Mathematical
< < Tables and Aids to Computation) -- the report itself, consisting of 23
< < mimeographed pages, was dated Nov 1949 according to the review.  
< 
< Just out of curiosity, what precisely were these modifications?  The way 
< ENIAC was programmed (setting thousands of switches by hand) makes it hard 
< to imagine that it would have been easy to convert it to a real stored-program 
< machine.

The ENIAC as implemented in 1945 was programmed by wiring it up.  It also
had some function tables that could be set by switches (as well as
various other switches).  The 1948 modification was to get a switch that would
send a pulse down one of 100 distinct lines depending on which two
digit code it read.  This was used to activate one of 100 different wirings
of the ENIAC that were then kept the same from program to program.  These
``machine code'' values were read from the function tables.  So at
this point, you had a read-only program store that you could set by
switches (although lost from your quotations, my original message did
specify it as a ``read-only stored program implementation.''

< < Also of interest was the second meeting of the ACM, which was held in
< < Dec 11 & 12 of 1947.  At this meeting, two papers were presented
< < whose titles suggest a relevance to this discussion (alas, I have not
< < seen these papers anywhere): ``General Principles of Coding, with
< < application to the ENIAC'' by J. von Neumann and ``Adaptation of the
< < ENIAC to von Neumann's coding technique'' by R. F. Clippinger.  The
< 
< I don't necessarily see the relevance.  I tend to think of "coding" as 
< "hacking" and all i can learn from the titles is that von Neumann had 
< a clever hack for something ... 

Nope.  Von Neumann is clearly talking about general programming
methodology and not some specific hack (go down to the math library
and take a look at volume 5 of the collected works of Von Neumann and
you will see the sorts of things he talks about under phrases like
``coding'').

< Since Eckert and Mauchly had the stored-
< program idea before von Neumann came aboard, and i would assume that 
< RFC was sufficiently closely involved to have known that, the titles would 
< say even less about the "modifications" ... 

What is at issue is not the stored-program idea (clearly Turing's paper
on the Turing machine had that as well as various other people throughout
the history of computing), but rather the implementation of this idea
using electronics.

I forget where, but somewhere I recall reading that either Eckert or Mauchly
claimed that Clippinger had independently discovered something that the
designers had always intended to leave open as a possibility. However, 
while the machine was at UPenn, there is no record of anyone trying out
machine codes -- the earliest layout of a machine code for an ELECTRONIC
computer that I have seen is Goldstine&von Neumann's EDVAC proposal.

< Somehow i find it hard to believe that Eckert and Mauchly lost all interest 
< in ENIAC, notwithstanding the move to BRL and work on the EDVAC.  If the 

Take a look at the Moore School lectures some time [reprinted MIT
press 1985; originally given summer 1946.].  Specifically, on page
490:
     A number of the students attending the special course had
     come under the impression that the main topic would be the
     construction and operation of the ENIAC.  Up to this point
     in the course the ENIAC had been mentioned only in passing
     [1 month + 1 week into the 2 month course], and then only
     to illustrate the concept of an accumulator.  Several students
     had asked the Moore School staff to include more ENIAC information
     in the lectures, but, according to Verzuh, they usually recieved
     the comment ``It is no good, so why talk about it.''  The dissatisfaction
     reached a climax on August 14, when Verzuh noted: ``During the
     day an inquiry sheet was passed out to the members of the class.
     There had been some boisterous denouncements of the manner in
     which the course was being conducted.  As a result of the inquiry,
     a discussion of the ENIAC is to be included in the next 12 sessions.

< big thing about the EDVAC was the stored-program idea (and given that 

Nope.  The big things were: SERIAL (no more parallel confusions); large
program store (no more scraping by with 20 registers); and finally
putting the program into that writable program store.

< BRL was presumably footing most of the bill for the two machines) then 
< it is unlikely that the change would have gone unnoticed; recall that there 
< was someone (Goldstine) whose job was to act as liason between BRL and 
< UPenn ... and Goldstine could hardly wait to let the world know what a 
< great idea EDVAC and its stored programs was ( ... the business with 
< the preliminary EDVAC report).

I don't recall saying Eckert and Mauchly didn't know about it.  Merely
that they didn't seem to have been involved in it.  At this time they
were also breaking their ties with UPenn as well.  Both of them left
UPenn in early 1946 to form their own company -- two years before the 1948
mods.

---- BOB (webber@athos.rutgers.edu ; rutgers!athos.rutgers.edu!webber)