Peter Paine
Wed, 8 Apr 1998 07:49:34 +0000

At 3:43 PM +0000 98/4/3, Christopher J. Vogt wrote:
>Kalman, did anything ever get published on the I-machine architecture?  I
>seem to find anything.

A UK company, Racal-Norsk, spent some 10 to 20 person years analysing and
writing up a formal specification of MIT Zetalisp.  The company went into
oblivion over a decade ago, but the copies of the spec gather dust under
the beds of some of the engineers who wrote it.

The document is exhaustive and rigorous and was the foundation of a working
implementation of Zetalisp which ran on Norsk Data superminis.  It draws
together as one whole the disperate contibutions and evolutionary stands
which made up MIT Zetalisp.  It was not published publically, it was
considered the crown jewels of the company.

I believe that any lispos work which derives from Zetalisp would find this
document extremely valuable (not that I am suggesting that this would or
would not be a wise course for lispos).

I understand that it is the only formal spec of lispm lisp - which is total
wasted were it is.  Despite incredulity from the US lisp machine
establishment the spec, company and developments were successful - perhaps
in part because some of the prime movers cut their teeth at the beginning
of the 80s with the CADR, besides having RMS and others on the project.
Anyway I suspect that in those days Lisp outside the US was seen only as an
sales operation.

The company evapourated whilst at full steam and a win (in a Lisp Machine
sense) when the joint venture which formed it fell out of bed.  At that
time, Racal-Norsk (I believe) owned the exclusive rights to MIT Zetalisp in
the UK (cheekily (falsely) Symbolics Genera products distributed in the UK
claimed that right) and R-N had a worldwide (ex USA) right to Zetalisp
distribution.  I suspect that neither Racal nor Norsk Data would even
recognise the word Lisp today, but if they have residual rights to MIT
Zetalisp, might there be a chance through that route to free Lispm Lisp
(though I would expect the licences were time limited, but I cannot

My copy of the spec is probably not complete, nor in digital form, other
are -  though their possessors have moved on to other pastures.

Of course the really valuable input to the lispos project could be the
dormant Lispm movers and shakers from MIT, Symbolics, Xerox, Texas, LMI,
etc.  A good listing of whose names can be found in the second Symbolics
world distribution released after the namespace was introduced (Genera 7.1
?), by inspecting the namespace instance, you can browse backwards into
both Symbolics and MIT's complete namespace data (someone didn't do a good
enough house cleaning before shipping).

Otherwise what about the simulation built by Symbolics from which to
develop the Ivory architecture, I remember hearing some Symbolics
heavyweights rave about this.  I understood that this worked as an emulator
which could model both sides of the interface between hardware and lispm
primitive instructions, and in an accessible object oriented manner.

Or is this heritage just so much old rope, and what is needed is new minds,
new ventures, new designs, ... just as Lispm/Lisp needs a new image, new
packaging, new customers/users and uses, perhaps escaping the ivory tower
and entering the mainstream.