My feeling about some "features" of Macsyma / Maxima is
that deserve to be anti-documented. That is, there are
some things that are so complicated that it is impossible
to document them, and so a better version, and its
documentation, should be provided. Unfortunately
this requires good taste and maybe even some level
of consensus. Most people don't want to look at,
critique, and reach consensus about someone else's
design.
For what it's worth...
I suspect that Wolfram's Mathematica book came about
this way
(1) Wolfram learns to use Macsyma
(2) Wolfram thinks he can do better
(3) With Geoff Fox and some others, writes SMP
(4) Tries to make money with SMP,
(5) SMP turns out to be a loser; start again
(6) Wolfram write a description of what a CAS should be
(7) He and others write a system to do that. Modify the
description as little as possible.
Documenting an existing system "after the fact" is
not so easy.
I'm going to try to put online a
paper from the 1979 Macsyma Users' Conference
"Macsyma's General Simplifier: Philosophy and Operation"
(10 pages, single spaced).
Maybe that will help some of you.
rjf