Unix History

[Henry Spencer]

> > Although AT&T won't admit it in so many words, they effectively
> > abandoned work on PDP11 Unix a long time ago.
> 
> Most new features made it to the PDP11...

But there weren't all that many new features.  A lot of what was new was
the performance work, which did *not* make it.

> > For example, although some of the SysV performance work wouldn't fit
> > on the 11, *some* of it would.  None of it was applied to PDP11 SysV.
> 
> The PDP-11 uses hashing to implement the sleep/wakeup facility in the kernel
> under System V...

It did under V7 too; this is nothing new.

> I don't know of any other System V performance work that
> was applied to the PDP11 (but then if it had been it would probably would not
> have helped a machine the size of a PDP-11 much.)

Nonsense.  Things like inode hashing make quite a substantial difference,
and are easy to put in, if anyone had bothered to *try*.  As for "the size
of a PDP-11"...  an 11/44 is fully the equal of a 750 on anything that
doesn't hit address-space problems, and an 11/70 approaches a 780.  The
neglect of the 11 was not because the machines wouldn't benefit from it,
but because AT&T had, as I indicated, effectively abandoned the 11.  (I
don't *blame* them, given how hard the address-space problem hits the
kernel, but they should stop lying about it.)

> > ... PDP11 SysV is really SysIV.
> 
> Not really.
> 
> >  The PDP11 SysV shared-memory stuff is also different from and
> > incompatible with the regular SysV version.
> 
> Right, but it is present, which is a difference from UNIX 4.0.  The IPC
> stuff was first released in UNIX 4.2.

I wasn't claiming that PDP11 SysV was identically equal to 4.00000, but
that it was largely one of the 4's, not really a 5.  (I should have been
clearer about this.)
-- 
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