RCS / NROFF under SYSTEM V

[Guy Harris]

> Nroff is (originally) an AT&T Bell Labs product;

Still is, unless it's considered an AT&T-IS product now.

> the System V version is somewhat improved from the version in 4.1BSD (4.2
> may have done some improvements also.

The 4.2 version is a minor change from the 4.1 version, which is in turn
still pretty much the V7 version.  They added some changes that I've been
told are necessary to make the "-me" package run (those are the lines
flagged with the famous Berkeley comment /* XXX */).  They also changed it
to keep the temporary "file" in memory.  This can be put into the S5 version
as well.

NOTE: the "INCORE" option for the S5 nroff is *not* what you want.  It's for
large VM machines which aren't running UNIX; it uses "malloc" instead of
"sbrk" and standard I/O instead of read/write.  At the time I was playing
with it, changing it to use "malloc" made it run *lots* slower - which is
the last thing "nroff" needs.  I can't speak for the standard I/O vs. UNIX
I/O difference.

Also, the INCORE option disables "compacted macro packages"; the S5 "nroff"
can be told to dump its temporary file with some header glop after reading
in a macro package, and then can reload its temporary file.  This may speed
up loading of macro packages, especially monsters like "mm", so you may not
want to eliminate it.  (The macro package needs to be hacked up to make this
work, so if you want to speed up loading of "me" be prepared to spend some
time on it.)

> The -me macros are a Berkeleyism.  I suppose they'd probably work ok on
> System V nroff.

They seemed to, after the /* XXX */ stuff was stuck in (I believe we were
able to format the Sendmail documentation at CCI).  I don't know about
vanilla S5 nroff, but you might give it a try.

> They should be available on a 4.*BSD system, probably in
> source form under /usr/lib/macros, even if you just have a Binary license.

They are, but the source is in /usr/lib/me (there is no /usr/lib/macros on
4.xBSD or V7).

	Guy Harris