28 Jan
2002
28 Jan
'02
1:48 a.m.
You said it! The "thing" about one-shots and delay lines is uncertainty.
Synchronous logic is rigorously predictable, which one-shots are not. Delay
lines, likewise, are not entirely predictable because the analog delays they
present varying time intervals. This may seem to be a minor point in light of
the many years during which those uncertainties were tolerated.
However, as time went on, large manufacturers (not just IBM) learned that the
rigor of entirely synchronous logic produced better results over the long haul
than the often more clever and smaller circuitry that asynchronous logic
brought with it, because with the asynchronism, the synchronous design left
the uncertainty behind as well.
Dick
----- Original Message -----
From: "William Donzelli" <aw288(a)osfn.org>
To: <classiccmp(a)classiccmp.org>
Sent: Monday, January 28, 2002 7:55 AM
Subject: Re: Netscape (was Re: PayPal = payola?)
Since when are
one-shots are a problem, they have there place.
Thinking back more; most timing was asynchronous that needed
a lot set/reset flip flops and delay buffers to keep things in sync.
It is tricks to save a gate or two that is the problem.
It is not "the IBM way". Back in the 1960s and 1970s, IBM strongly tended
to have everything clocked, sometimes using rather impressive circuitry.
The clock generator of an S/3 is astoundingly complex, and gets around
some of the uncertainty involved with one-shots and delays.
I suspect a hard code sequencer.
No, it was a real microcode machine. It could run a set of complicated
diagnostics on a disk, without bothering the channel. Real front panel, too.
William Donzelli
aw288(a)osfn.org