What should GNU run on

[Keith D Gregory]

It seems that some thought should be given to history here.  As I recall from
first reading "The GNU Manifesto", GNU is not to give any notice to such
"insignificant architectures" as the 8086 (as I recall, the original wording
was not so overtly inflammatory; one had to read between the lines).  As a
result, it seems that most of the code is currently written for 68k or VAX.

So, now the .*86 line has significant power, has mostly shaken the constraints
of its segmented architecture, and is the most available and most inexpensive
platform in existance.

But is it going to be that easy to throw away history, such as GCC et al?  If
GCC was written to produce optimal code for the VAX/68000 (dare I say PDP-11)
style architecture, will it transport well to the Intel architecture?  Is a
project currently underway (or finished) to do so, or has anyone simply made
a [non-optimal] code generator using the current compiler?

********************************************************************************

I will agree, however, that a 386 (or 486) is probably the best platform, simply
due to the fact that it is such a widespread platform.  Perhaps a Mac would also
be a contender, although it seems to me that one could go overboard trying to
usefully use its user-interface (multiple sessions would be nice, tho' ... it's
something that A/UX is missing, unless you want 3 pty's per session).

I think that SPARC, VAX, Sun-{2,3}, &c are not realistic.  If you have a uVax
that isn't being used for anything else, more power to you.  Same with a Sun.
Most people, however, aren't going to be able to convince their local MIS dept
to turn over that old 780 for use as a GNU system.  Nor are most people going
to buy a SPARCStation, even at $9995, simply so that it can sit beside their
AT-386 or Mac-2, and run GNU (OK, to keep a religious war from starting, I will
simply say that _I_ wouldn't, although I might think of buying another drive
for my Mac-2).

-kdg