The resurrection of Joe Marshall's home page prompted me to browse through his collection of LMI K-Machine code. It's not just Lisp source code, though; there are a number of interesting text files. Here are some bits from k-lisp.txt.
On supporting both ZETALISP and Common LISP code on the system:
Where Common LISP specifies that "it is an error" to do some particular thing, ZETALISP allows such a thing. Example "It is an error if the array specified as the :displaced-to argument does not have th esame :element-type as the array being created." ZETALISP allows that and "does the right thing" (which, actually, it shouldn't as anyone using this kind of bagbiter is a loser anyway.)
On limiting access to system source code:
It may be true that we treat our customers like children, but unhappy customers generate less income than bug reports and pull down valuable programmers. I would rather piss off a thousand RMS's than one MCC.
Abstraction is the most powerful tool to reduce the complexity of a system and make it understandable. This is an amazingly simple concept that almost everyone ignores totally. They lose because of it.Fun stuff.