<!DOCTYPE html>
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
</head>
<body>
<div class="moz-cite-prefix">On 1/3/26 11:01 PM, Peter Gutmann via
cryptography wrote:<span style="white-space: pre-wrap">
</span></div>
<blockquote type="cite"
cite="mid:ME2PR01MB3650736ABDA22B0C1617C849EEB9A@ME2PR01MB3650.ausprd01.prod.outlook.com">
<pre wrap="" class="moz-quote-pre">Probably time to post a link to Lawrence Wilkinson's heroic resuscitation of
an IBM 360 model 30, which included designing his own 6802-based channel
peripheral by reverse-engineering things from the circuit diagrams and
microcode:
<a class="moz-txt-link-freetext" href="https://www.ljw.me.uk/ibm360/Saga/">https://www.ljw.me.uk/ibm360/Saga/</a>
The console of that machine has been preserved:
<a class="moz-txt-link-freetext" href="https://www.cs.auckland.ac.nz/historydisplays/FourthFloor/MainFrameComputers/System360Model30.php">https://www.cs.auckland.ac.nz/historydisplays/FourthFloor/MainFrameComputers/System360Model30.php</a></pre>
</blockquote>
<p>So the polished up console still exists, but the working machine
he resurrected is long gone?</p>
<p>Hmmm.</p>
<p>I'm sad that it is gone, but it also sounds like it was a hellish
collection of engineering: Impressive it was once worth designing,
building, debugging, powering, and <i>servicing</i>…and also a
good riddance. Reading microcode from tarnishing silver ink off of
punch cards held in place by compressed air‽‽ The 6802 he used to
emulate missing IO hardware is itself now a museum piece that must
be decades out of production. (Or is it? I think a version of 555s
is still in production.)</p>
<p>Clearly early automatic computing was so valuable that it was <i>worth</i>
such expensive Rube Goldberg contraptions. But look at that
machine and extrapolate forward…and the whole thing was clearly
not scalable, obviously there was only a limited future for the
industry. But we didn't extrapolate that technology, thank God.</p>
<p>And now I look at LLMs, again rather Rube Goldberg on the inside,
impressive in what they can do, and when one extrapolates forward,
also not scalable. But apparently even more compelling than the
then-new ability to run a few expensive, slow COBOL programs, as
we are full speed ahead scaling up this new technology anyway.</p>
<p>Weird being so old as to see such dynamic range.</p>
<p>-kb</p>
<p><br>
</p>
</body>
</html>