Issue: PATHNAME-SUBDIRECTORY-LIST (Version 2)

[masinter.pa@Xerox.COM]

I'd suggest we pare down the proposal PATHNAME-SUBDIRECTORY-LIST and allow
practice to follow: allow a list to be a filler of a pathname. Pathnames
are "relative" in an implementation dependent manner (maybe they're
relative to ROOT). And some implementations may accept and return special
keywords or lists of (keyword string) instead of components for the
DIRECTORY component of a pathname, we recommend:

:ABSOLUTE (allowable at the head) meaning to start at the "root" of the
file system

:WILD (allowable anywhere)
:WILD-INFERIORS
:UP


For example, in Unix systems, "/a/b/c/" might parse to (:absolute "a" "b"
"c") while in VMS "[a..b]" might parse to ("A" :wild-inferiors "B").

I think we should say that conformal programs that deal with pathnames must
'expect' those results back from (PATHNAME-DIRECTORY ...) but not require
implementations to return simple strings, instead making it "advisory". 

This allows us to write a standard that talks about Unix even though there
isn't an ANSI standard for Unix, and let users know what to expect, even
though it isn't absolutely required for conformance because we can't
imagine all of the possible keywords that might map onto host file systems.

What do you think?