@amoroso

I tried to make this point to the marketing dept in about 1987, when we were finishing up the Joshua expert system language.

You could pass objects between Macsyma and Joshua, to get expert system advice on math.

You could pass objects between Joshua and Statice to get smart object databases.

... and so on.

I even went so far as to write an example system that scraped weather data from the net, stored it in a Statice database, and then used a (cartoonishly simple) Joshua expert system to make short-term weather predictions (temperature, sign of the time derivative of smoothed barometric pressure, that sort of thing).

By "short-term", I mean it would predict rain about the time you could look out the window and see it's raining. 😄

But it was *really* hard to get people to understand the power you had when all those systems worked in the same address space!

The idea survives today only as an antique Symbolics marketing video; see about 11:15 in the video below.

https://www.youtube.com/watch?v=6uPwQuxjgQo

@amoroso

If you were at Symbolics, you could inspect it *all* the way down, including the microcode. 😄

But given the microcode word was pretty wide, that meant lots of micromachine parallelism. That got really arcane, really fast. I never learned much about it. It was never released to customers, because it would take a year to get any good at it, and we could never have supported it anyway.

And with the I-machine, all that was in ROM on the chip and not changeable anyway. (Probably a very good move.)

Too bad the S-machine apparently was cancelled the day it taped out.

1+ more replies (not shown)