27 Feb
2015
27 Feb
'15
8:41 p.m.
On 2015-Feb-26, at 7:51 AM, Evan Koblentz wrote:
Thanks for the link, the articles were very interesting.
However:
Fuel for the
eternal fire: Goldstine seems to be pretty definite in attributing the stored-program
concept to Johnny.
Prominent science historian Tom Haigh recently led a team of researchers to get to the
bottom of this. Their work resulted in three recent academic papers
and a new book coming soon, in which they clearly found that ENIAC was up-and-running in
stored-program configuration a short time before the Manchester and any other computers:
www.eniacinaction.com.
The authors of the articles carefully avoid declarations of ENIAC as a stored-program
machine. They separate out 3 aspects from the "First Draft of a Report on the
EDVAC" (the "modern code paradigm" as defined by them, the "von
Neumann architecture paradigm", and the "EDVAC hardware paradigm") and show
how the 1948-modified ENIAC (to provide it with an instruction-set-based programming
capability) implemented the "modern code paradigm". While that was a couple of
months before the Baby, it is not the same (by the authors own definitions and intent) as
making it a stored-program machine.
In modern terminology, I would suggest the 1948 ENIAC would be classed (as far as it can
be) as Harvard-Architecture or a form of modified-Harvard-Architecture machine.
John v.N. showed up after this, wrote down what he
learned there, and sat back while everyone started calling it the "von Neumann
architecture".
Von Neumann was around the project years before the 46-48 modifications. He produced the
Draft Report on the EDVAC - from which those modifications were conceptually derived - in
1945. As presented in the articles, the modifications were von Neumann's idea and
implemented in consultation with him.