[TUHS] update pdp11 system five?

[Clem Cole via TUHS]

below

On Wed, Dec 17, 2025 at 10:22 AM Jacob Ritorto via TUHS <tuhs at tuhs.org>
wrote:

> pdp11s are getting so popular these days that now seems a great time to
> perform the stunt of running SYSV well on them.

That will be a great deal of work, more in a minute.

>   Seems pretty normal to the adepts on the list who were there when it
> originally happened, but it’s kind of a niche mind-blow to newcomers that a
> PDP could run SYSV.
>
Be careful, System V Release 1 (SVR1) ran on both DEC PDP-11 and VAX-11/780;
the PDP-11 was the reference architecture, while a VAX-11/780 port was also
offered as an different UNIX implementation from 4.1BSD [the new memory
management code for the Vax, was considered much cleaner - I believe much
that came from the early Reisner Vax porting effort].

SVR2 (second release): The official "porting base" / reference architecture
was the DEC VAX-11/780. With SVR2, AT&T's USG was shifting focus away from
the PDP-11 as the target platform, though legacy code likely remained.
That said, I don't remember a PDP-11 release for SRV2

SVR3 (third release): The official "porting base" / reference architecture
was the WE 32000-based (called the 3B2, IIRC).

With SVR3. Stream-based networking using the Transport Layer Interface
(TLI) was added, but an IP/TCP stack was not provided.  A number of folks
such as Lachman offered them

SVR4 (fourth release): The official "porting base" / reference architecture
was the Intel 386/486 family

>
> But currently I think it’s missing networking.
>
It depends on which flavor of System V you start.

>
> So a couple questions:
> Is it possible to build newer SYSV releases to run on pdp11?

The SVR1 boot tape (which you can find in the wild, although the
provenance might be shady.  I've never seen a PDP-11 boot tape for any of
later release (SVR2/3/4)



>   Like specifically SVr4?
>
Not out of the box. Your first big hurdle is going to be a PDP-11 compiler,
much less one that self-hosts [modern C compiler will not fit in the 64K
(even with 64K data/64K text for the separate I/D processors).


> If not, is this feasible to achieve and would anyone be interested in a
> project to make that happen?
> If that can happen, the stock SVr4 networking code build might be a breeze.
>
I doubt it will be a breeze.  As others have pointed out, 2.11BSD uses the
DEC overlay code from the DEC PDP-11/V7 release, and managed to cram a much
larger kernel into memory.  But note that their IP/TCP stack runs in a
separate address space (supervisor mode).

>
> Apologies in advance if this has already been broached and completed
> and/or if this is a fool’s errand due to size constraints.  And yes of
> course the easy thing to do is just run v6 or BSD.  But then we miss out on
> the mind-blowing facet :)
>
Well, of course, V5/V6/V7 will all "just work".  If you are willing to give
up a networking API, you can also run SVR1. But if you really want a more
modern take on UNIX on a PDP-11, 2.11BSD runs exceptionally well and has a
reasonably present group of people who are creating bug fixes.  But I will
warn you: Steven Schultz (who leads this effort) makes it clear that the
goal is not to recreate 4.2/4.3/4.4 on the PDP-11. Instead, 2.11BSD is to
allow a modicum of modern tools to run there still. You can look for it at:
2.11BSD <https://www.tuhs.org/Archive/Distributions/UCB/2.11BSD/>


Clem