The spec for a language is harder to define than the compiler for it. The spec for Lisp is written in Lisp. And it's about half a page of Lisp code. And it's executable, which is not usually true with specs.
This is why reading the Lisp spec feels like reading "the Maxwell's equations of software" (to quote Alan Kay). It's a whole other fundamental model of computation, every bit as fundamental as a Turing machine, but much more practical. Nobody builds serious computing engines out of pure Turing machines, but people do it all the time with McCarthy machines.
The LISP 1.5 Programmers manual! Love this one. Your 2) is what Alan Kay called "Maxwell's Equations of Software".
And you're right, there's absolute magic in old computing papers. They were pioneers. I wish we could see more of this pioneering spirit combined with todays silicon and ease of international communication.
Years ago I made it partway through this early LISP manual co-authored by a young John McCarthy before his move to Stanford: http://www.softwarepreservation.org/projects/LISP/book/LISP%...
I remember thinking:
1) They did not mess around in the 1960s. No "Hello World," no promises, no sales pitch. Just an in-yer-face "Here's how we do things downtown"
2) They defined a language in terms of itself. I still don't get all that implies but, damn, they can do that?