AT&T "nth Edition" vs. "Release n"

[Guy Harris]

>(The problem with talking about `System V' is that the four `Release N's
>are completely different, and a `Release N' for machine A is completely
>different from a `Release N' for machine B.

Well, not *completely* different, although it's far more different than
it needs to be or than it should be.

>The region VM was in the VAX versions of VR2 and VR3, and maybe even VR1,
>if my faulty memory is giving the right answers.

SV"R1" (I don't think they ever actually called it "Release 1"; back
then, they may have still thought that the successor to "UNIX
5.0"/"System V" would be called "UNIX 6.0" or "System VI", rather than
"System V Release 2"; perhaps the fact that "System VI" came with, well,
"vi" made them nervous about the name :-)) had a swap-based VM, with
largely V7ish (or maybe 32Vish) underpinnings.  System V Release 2 on
the VAX did as well; S5R2 Version 2, or whatever it was called, had the
region VM.

The 3B2 paging S5's had it as well, as I remember - at least S5R3 does -
albeit with changes for different hardware that pervade nominally
machine-independent portions of the kernel far more than they need to. 
"Fixed in 4.0", with the introduction of the SunOS VM system.

>But the System V source trees for different platforms were entirely
>separate.  Keith Bostic once had an argument with some USDL-type who
>claimed that merged trees were impossible, even though 4.3BSD-tahoe
>had already shipped!)

Yes, there is that rather gaping blind spot in the vision of some of
them; some USDL type (perhaps the same one as Keith spoke with) was
quite shocked to hear that SunOS 3.0 for the VAX was built mostly from
the exact same source tree as SunOS 3.0 for Sun-2s and Sun-3s (the
differences being things such as the assembler, parts of the compiler,
machine-dependent parts of the kernel and libraries, etc. - the bulk of
the kernel and user-mode stuff was shared).