Issue: TRACE-FUNCTION-ONLY (Version 4)

[David A. Moon <Moon@STONY-BROOK.SCRC.Symbolics.COM>]

I favor version 4 of this discussion, and I favor the proposal named
TRACE-CLOS::TRACE-FUNCTION-SPECIFICATION (it's really supposed to be
named TRACE-FUNCTION-ONLY:FUNCTION-SPECIFICATION, isn't it?), provided
the Adoption Cost section is edited to point out that this is an
incompatible change for several implementations that already extend
Common Lisp by defining the meaning of a non-symbol as a subform of
TRACE.  In particular, at least Symbolics Common Lisp, Xerox Common
Lisp, and CMU Common Lisp define a list as a subform of TRACE to
be something other than a function-spec, and probably a lot of other
implementations do something like this.

I favor the change even though it's incompatible, because TRACE is part
of the environment rather than part of the programming language, meaning
that compatibility is less of an issue, and because all of the existing
extended syntaxes of TRACE that I am aware of are horrible.