[TUHS] porting to different systems, Bootstrapping UNIX - how was it done

[Clem Cole via TUHS]

On Tue, Mar 24, 2026 at 12:34 PM Paul Winalski via TUHS <tuhs at tuhs.org>
wrote:

>
> The original PDP-10 did not have virtual memory capability.

right KA10/KI10s

> BBN designed their own paging hardware and wrote their own operating
> system called TENEX.

See Gunkies: https://gunkies.org/wiki/TENEX


> DEC did implement paging hardware for the KL10 processor, which was
> used in the DECSYSTEM-20.  But the paging system differed from the BBN
> design and TENEX would not run on the DECSYSTEM-20.

Not quite right.  At least one KL10-E shop modified the microcode, and I
believe a few small changes to TENEX were also made.   Those sites were not
the only ones to make changes; MIT's MC machine  [1090]  was a KL10-PV,
sometimes the model B. Since ITS was created for the older MIT modified
KA10's 18-bit addressing and had its own unique paging scheme, Tom Knight
(one of the MIT engineers) made hardware modifications to the KL10-PV's
paging logic/microcode [*a.k.a.* "extended addressing].



> By then TENEX was extremely popular in DEC's PDP-10 customer base.  DEC
> bought the rights to
> TENEX and ported it to the DECSYSTEM-20, the main changes being to cope
> with the different paging system.  This port was TOPS-20.
>
Right.

>
> TOPS-10 would run on some DECSYSTEM-20 models,

KL10-D and Es were the same core CPU.   TOPS ran on all of them fine, it
just loaded different microcode as I explain previously.



> in particular those using the KS10 CPU.
>
Actually, not until much later.  The KS10 was sufficiently different from
the KL10; TOPS-10 had to be modified.  The 2020 running TOPS-20 was
released in the first part of 1980.  TOPS-10 version 7.01 was not released
for another 2 years (1980).  In that release, support for KS10 was added,
but the primary addition was to support Master/Slave processing on a dual
KL10-E.   That said, in the same timeframe, TOPS-20 Release 4 came out, and
it supported "fully symmetric" operation on a dual-processor KL10-E (though
there were limits, as true "shared-everything" symmetry was constrained by
the hardware itself).