Computing Pioneer Dies

Dave Wade dave.g4ugm at gmail.com
Sat Nov 11 04:13:55 CST 2017


> -----Original Message-----
> From: cctalk [mailto:cctalk-bounces at classiccmp.org] On Behalf Of Jason T via
> cctalk
> Sent: 11 November 2017 04:16
> To: General Discussion: On-Topic and Off-Topic Posts
> <cctalk at classiccmp.org>
> Subject: Re: Computing Pioneer Dies
> 
> > I'm harly a member of the "ENIAC/Mauchlyite crowd" (in fact, I used to
> > not have a good impression of them at all), but I thought Haigh et al
> > made a pretty good case.
> 
> Here's Prof. Haigh speaking on ENIAC at VCFMW last year:
> 
> https://www.youtube.com/watch?v=0O0pKPzJjEY

Thanks for that, I haven't (yet) watched the video (its an hour long but looking forward to it) , or indeed read the whole book but I have read the papers by Crispin Rope and also listened to a similar talk by Mark Priestley another co-author of the book, after which we chatted at length about all this. Whilst they make a good case they tend to skirt over the fact that ENIAC ran its stored programs from what in effect was a ROM. ENIAC just didn't have enough "RAM" (technically it just had registers) in which to store data and programs. So they re-purposed the function switches which, if I remember properly, were originally intended to be used for storing polynomial co-officiants or other parameters, I guess much in the same way as we us an input variable in a modern calculation, and used them to store a program.

To do this required them to, in effect "microcode" a EDVAC like machine onto the ENIAC hardware. To me this was brilliant thinking, given that at the time they started no one had built such a machine, and of course so long as they included conditional branches, which they did, the machine will be Turing Complete. They then went on to use it to solve many problems, and by using the function switches to store they reduced the re-programming time from weeks to hours. Less for a small program.

Of course modern computing without RAM is inconceivable and what the Baby/SSEM had was a program in RAM. You don't need to have the program in RAM for Turing completeness, but could anyone envisage modern computing without RAM. Note the concept of RAM wasn't new. In effect it’s the basis of a Turing machine which has an infinite tape which can be written, moved, erased and re-written is an infinite RAM. Nor was the implementation using a Williams tube long lasting, but at the time it was novel, simpler and cheaper to implement than Mercury Delay lines (EDSAC at Cambridge contained around a Ton of mercury), and IBM licenced the technology for the 701. When core became available it quickly replaced other storage technologies (except in the UK where we continued to use delay lines, but using torsion wire note Mercury, I assume to avoid the expense of licencing core.

Dave
G4UGM
"Baby" Volunteer @ MSI Manchester





More information about the cctalk mailing list