>So, I have doublings every two years, every year, and every 18 months,
>all offered as statement of Moore's Law.
>
>I suspect Moore originally said doubling either every one or every two
>years, and revised this to doubling every 18 months. But it would be
>interesting to know the truth.
Moore's "Law" gets different versions because it's not as much of a "law" as
most people claim. Over the last 35 or so years, the doubling rate has
varied between one year and two years, with 18 months a decent long-term
estimate. I believe the "law" was originally stated as every year, because
it was doing just that for several years in the 60's. It was revised to 18
months in the late 60's. The 70's and 80's had several periods of about 2
years, and at least once in the late 80's Moore changed it to every two
years. Now, I think, it's around 18 months again.
Moore himself is quite upfront that it's an empirical observation and has
varied over the time period. In Wired recently he said that the law will
probably hold for another 10-12 years but doubts that it will hold beyond
that. Beyond that, we will have to move to a radically different system than
wiring and transistors photographically etched on silicon. There's no
particular reason that this other technology (which might not exist then)
will have the same emergent phenomena as current chip technology.