geof%MIT-BORAX@Mit-Multics@sri-unix (12/13/82)
I beg to remind you all that lisp was first written on a machine with less than 64k of memory, and ran in ~32K of memory for the first 5 years or so of its life. Further, I have seen wonderful and useful lisp implementations on the PDP-11, which has a 64K address space. Look, you're not going to get MACLISP or a Z-LISP on an 8080 class machine; that's clear. And the smaller your memory (and as heap's go, 64K is peanuts) the more time you'll spend garbage collecting. But given that only one person uses the machine at a time, if you're clever, you can put together a pretty useful lisp that runs in 64K of memory. I agree that in this day and age it is better to go for a system that does not require one to count bytes. But I really don't believe statements to the effect that a useful lisp system just will not fit into 64K of memory. - Geof Cooper
RG.JMTURN@Mit-Mc (12/16/82)
Date: Monday, 13 December 1982 13:43-EST
From: geof%MIT-BORAX at Mit-Multics
To: info-micro at BRL
Re: 64K lisp
I beg to remind you all that lisp was first written on a machine with
less than 64k of memory, and ran in ~32K of memory for the first 5
years or so of its life. Further, I have seen wonderful and useful lisp
implementations on the PDP-11, which has a 64K address space.
I suspect people's definitions of "useful" are different. To someone like me,
who uses Lisp on an everyday basis, MacLisp is the borderline of acceptability.
Remember also that a PDP-11 is not an 8-bit machine, so you get more bang for
that 64K,
Look, you're not going to get MACLISP or a Z-LISP on an 8080 class machine;
that's clear. And the smaller your memory (and as heap's go, 64K is peanuts)
the more time you'll spend garbage collecting. But given that only one
person uses the machine at a time, if you're clever, you can put together a
pretty useful lisp that runs in 64K of memory.
Lisp Machines are single user systems, and they don't cut it with 64K either.
I don't consider the 100-150 function Lisps advertised to be useful. They
tend to lack key functions that do not permit closure over the language.
This is usually most chronic in the string code.
I agree that in this day and age it is better to go for a system that does
not require one to count bytes. But I really don't believe statements to
the effect that a useful lisp system just will not fit into 64K of memory.
It depends. I'd buy a CP/M Lisp for different reasons than you. You'd
probably buy it as a toy, to play around with Lisp. I'd want it to replace
the primary system/applications language, which means I need a compiler,
interface to the hardware, and to CP/M, as well as an Mince interface
and a good error handler and allocation scheme and garbage collector.
That product just doesn't exist yet.
James