Peter Deutch's PDP-1 LISP ran in a 8K machine. The kernel of 
Smalltalk-72 could also, but we usually used about 32K. FORTH can run 
in very small environments. A FORTH-like Smalltalk would be a nifty 
little system. Larry Tesler's Tinytalk could do quite a bit in a very 
small environment (I don't remember just how small it could go). One 
of the keys here is that the pointer space is small, and the object 
space is even smaller......

Cheers,

Alan

At 6:21 PM -0700 9/12/02, Ned Konz wrote:
>On Thursday 12 September 2002 02:17 pm, Travis Griggs wrote:
>>  8K is, well, small. I'm just not seeing it happen with Smalltalk,
>>  even Wisp. The 'size' of the Wisp VM (built for x86linux) is 27026
>>  bytes. There's probably still some Palm related stuff we haven't
>>  hunted down and removed which will make it smaller, but we also
>>  plan to add primitives for socket support and that will undoubtedly
>>  offset that. An unstripped version of the image is about 10K.
>
>Though for many embedded projects (probably including this robot),
>both file and socket I/O is irrelevant. After all, most small
>embedded projects don't have an OS and so don't have files or
>sockets.
>
>I'd just as soon have the memory back.
>
>Of course, having i2c, SPI, serial, EEPROM, Flash, and other routines
>is a must for many smaller systems. But these can be loaded as
>libraries if you do register and bit access properly.
>
>Processes are not absolutely essential; I find that state machines are
>more useful for small systems. And these can be done entirely in
>Smalltalk.
>
>One missing part, though, is interrupt handling. A small system that
>can't handle interrupts is often useless for any but very simple
>"toy" programs.
>
>--
>Ned Konz
>http://bike-nomad.com
>GPG key ID: BEEA7EFE


--