[net.arch] Again, Why Not vfork

sbs@valid.UUCP (07/11/86)

> > UNIX is so good because it is portable.  SVID is supposed to go a long
> > way to make it MORE portable.  fork() is not-so-portable.  vfork() is
> > portable.  Why not use vfork() rather than fork()?
> 
> Vfork is not as portable as you think -- multiple processes within the
> same address space are hard to implement on some types of multiprocessor.
> By contrast, fork can be done well on any machine with a decent MMU.
> Portability does not mean that Unix has to run on every machine, no matter
> how bizarre.  Since fork works fine on everything from an IBM PC (retch)
> to a Cray 2, where's the problem?
> -- 
> Usenet(n): AT&T scheme to earn
> revenue from otherwise-unused	Henry Spencer @ U of Toronto Zoology
> late-night phone capacity.	{allegra,ihnp4,decvax,pyramid}!utzoo!henry

I'll second the emotion that "Portability does not mean that Unix has to 
run on every machine, not matter how bizarre."  Don't fool around with
Michael Rodent hardware -- it's not worth it.  By the time you've gotten
your multitasking 8080 system running with vfork(), we'll all be dreaming
the Great Bit Dream...

As to the comment that "UNIX is so good because it is portable": Nuts.
UNIX was a good system when it ran only on the 11; people were willing
to make the effort to port it.  In recent years, the portability cart has
moved around to the front of the UNIX horse, propelled no doubt by the
noxious exhalations of the UNIX-as-panacea promoters.  Get it right.


S.