Binary Compatibility 80286

[Steve Dyer]

I'm afraid that all the folks who are meta-flaming at Bill Gates'
comments regarding binary compatibility in Chuck Forsberg's
original message are really missing the point.

Face it: the software world operates on the distribution of binary images.
Source is either non-existent or extremely expensive.  It matters a LOT
to software developers that they might need a different binary distribution
for XENIX 86, PC/IX, VENIX, etc., even though they all support a single
microprocessor family.  One of the "strengths" of MSDOS is that the same
binary image runs on all sorts of machines which hew to a certain standard.
But, at this point in time, microprocessor-based UNIX systems are fragmented,
with no standards for system call numbering or invocation, no standard for
a.out format and loading, and so on.  Naturally, using 286 extensions would
be incompatible with the 8086, but there is no reason why all 286 UNIX systems
could not be binary compatible (with backwards compatibility for small-model
binaries.)

Of course, this is only meaningful within a microprocessor family like the
80x86 or 680x0, but it is nonetheless important.  Good software development
requires the existence of a "critical mass"; a certain market size which can
guarantee returns on one's time and money.  One larger homogeneous market
(i.e. all 8086 and 80286 UNIX systems) is far preferable to having to approach
a fragmented collection of smaller markets which might not be worth supporting.
-- 
/Steve Dyer
{harvard,seismo}!bbnccv!bbncc5!sdyer
sdyer at bbncc5.ARPA