In defense of DECUS C

[Jerry Leichter]

    Date:     Sun, 16 Oct 83 2:12:08 EDT
    From: Doug Gwyn <decvax!ucbvax!gwyn at BRL-VLD.ARPA>
    Subject:  Re:  need info: %r in printf
    To: decvax!yale-com!leichter at BERKELEY
    Cc: Unix-Wizards at brl-vgr.ARPA

    Just for the record, you CAN get a more recent definition of C; just
    buy the appropriate UNIX System V manuals.  You don't need a license.
    The language is much nicer these days than it was at the time of 7th
    Edition UNIX.

    I obtained the DECUS C compiler and was not at all impressed with it.
    To give the audience an idea of how its developers viewed C, the DECUS
    compiler was WRITTEN IN MACRO-11!  Most of the emphasis in its library
    support was on ways of accessing the baroque DECisms, especially bad
    in the RSX-11M version.  You would think that "Software Tools" would
    have had a larger impact..

(I tried sending a copy of this reply directly to Doug Gwyn, but it didn't
get through...Doug, I'm afraid you'll have to accept a public response only.)

The Unix System V manuals are expensive, quite recent, and hardly widely
available.  When I say "quite recent", BTW, please bear in mind that DECUS C
was written at least 5 years ago, and has changed relatively little:  Some
things for which hooks were left earlier - like floating point - were added,
but there have been no big, structural changes.  Hence, the compiler reflects
C as it was at about the time of K&R.  Also be aware that the compiler,
library, and everything else in the distribution were written entirely as
"midnight projects" by a small number of people (about 3).  (There are a lot
of contributors, but the compiler itself is mainly Dave Conroy's, with some
additions by Bob Denny and Martin Minow, while the library is mainly Martin
Minow's.  I have made some small contributions here and there.)  None of these
people were paid for this work.  They also eventually got a bit weary of
it. I can well understand why they don't want to bother to keep up with every
little of whimsy of Unix's.

As to the fact it's written in Macro:  Just what would YOU have written the
FIRST C compiler in?  Remember, there is NO existing C compiler to use:  You
can't use Unix because a) you have no access to a Unix system; b) you'd have
to go through an incredible amount of pain to move stuff to any other system
anyway; c) even if you had access, you'd have to be very careful about
avoiding license violations.  (EVERYTHING in DECUS C is public domain.  This
is quite deliberate, and is a carefully enforced policy.  Nothing makes it
into the distribution until we've checked and made sure that the contributer
really has the right to distribute the thing publicly, and has given up his
own claims.)

The eventual goal was to re-do the LIBRARY in C, and if you get the most
recent distribution tape - sent to DECUS late this summer - you will find a
good part of the library, in C, on it.  (In fact, that part moves quite
nicely to a 68000.)  It's mainly inertia, lack of time, other interests, and a
concern for efficiency in SOME of the library routines that keeps them in
Macro.  (Besides, the first rule programming has GOT to be:  If it isn't
broken, don't fix it.)

As for the compiler:  If you want to re-write it in C, you have the blessing
of everyone concerned.  Bootstrapping is a nice theory, but takes a LOT of
work, which would be hard to justify since DECUS C is already portable
among the systems it was meant to serve; PCC and other implementations are
just fine for "everything else".  (I'll address the general Macro vs. C
issue below.)

Yes, a lot of the library has to do with interfacing to DEC operating systems.
That's life.  Those operating systems have good sides and bad sides, as does
Unix; but the fact of the matter is that a lot of people use them, and with
DECUS C around they can now use C with them.  If your religious beliefs, lead
you to feel that people using DEC OS's somehow don't "deserve" C, you are
certainly free to do so; but we will ignore you as a fool.

Finally, as to lessons from "Software Tools":  Just what lessons is one
supposed to learn?  Many of the Tools are part of the DECUS C distribution,
in entirely new implementations (public domain, remember?).  They are written
in C, not Macro.  That has little if anything to do with what the compiler and
library are or ought to be written in.  I will say, however, that the reli-
gious anti-assembler bias I see all around me is absurd.  A well-commented,
well-written assembler program - which DECUS C is - is MUCH easier to under-
stand, maintain, and change than a poorly commented, poorly written C program.
Quality of writing, documentation, and commenting are much more important
than source language.  A program written in C is not somehow given special
divine "understandability" properties.  How many people in the world really
understand PCC - including Yacc?  From the comments I see in unix-wizards,
very few.  (I have no idea, BTW, whether PCC is poorly written; I do know that
most Unix code I've looked at is poorly commented, if commented at all, and
the low level of documentation quality for typical Unix programs is all too
well known to everyone except "wizards".)

						-- Jerry
					decvax!yale-comix!leichter
					leichter @ yale