NVRAM resuscitation (Was Re: SPARCstation 20 with SCSI2SD)

[Jim Manley]

DEC had some employees with clearances all the way up both primary sides of
the classification ladder, General Service (GENSER, which includes some
"black" programs), and Special Compartmented Intelligence (SCI, which has
its own alphabet soup, including other kinds of "black" programs).  They
needed this because magnetic core memory was used in PDP era systems that
processed up to the highest classification levels deemed prudent to not
require completely manual handling (including typewriters and carbon paper,
at least until pressure-sensitive carbonless paper was invented).

That included communications processing systems that, when the hardware
reached end-of-life (which happened to coincide with the upgrades made for
Y2K), the software was wrapped essentially in virtual machines that ran on
current-technology hardware and OSes.  The cost of porting the original
code would have been horrific because it had been mathematically proven to
be correct, and introduction of a single bug could not be tolerated.  Long
after PDP hardware had become commercially extinct, but the government
never throws anything away, DoD was still paying DEC lots of pretty pennies
to maintain a special secure version of the RSX-11 OS for feature
enhancements until the wrapping could be performed.

Core was so expensive that it was economically necessary to repair it,
rather than just replace it, especially in overseas systems where supply
lines were tenuous, at best.  As you might have heard, core never forgets -
at the Computer History Museum, when we resurrected our 1960-vintage IBM
1401, every bit of the auto parts database that it ran when taken out of
service was still intact over 30 years later.  That meant that military
commissioned officers had to escort double-wrapped/sealed/authenticated
packages containing such core devices all the way from almost-literally
Timbuktu back to DEC.

Plus, because of two-man rule handling requirements, two people with the
necessary clearances had to keep it in their presence when it wasn't
secured in a vault, and it was signed in and out every time it moved or
changed hands.  One of the benefits of volunteering to do this when flying
space-available on leave was that such couriers got to get on military
aircraft before anyone else, so we got first choice on seats.  Well, it was
as "choice" as it gets when often on tactical aircraft with jump seats used
by special forces and conventional paratroopers.

Civilians would describe the seats as ballistic nylon material stretched
between round aircraft-grade aluminum tubes ... aka a high-speed, low-drag
cot ... with nylon webbing, similar to seat-belt material, cross-woven
vertically/horizontally, with gaps equal in width to the webbing, hung
behind serving as a "back rest".  Military aviation seat belts were
thoughtfully provided passing through the webbing and secured to the inside
bulkheads of the fuselage ... mostly to make it easier for recovery crews
to gather the bodies in case of a crash ...

So, yeah, there's a whole world of physical security associated with NV
memory devices, alone, even if the technology has changed.  BTW, it's not
physically possible to reliably degauss every bit on rotating magnetic
storage media with a flux density higher than about that of a 1.44 MB
floppy disk, no matter how strong a field is produced.  It has to be
physically reduced to dust smaller than a specified particle size, or
incinerated.

On Wed, Nov 28, 2018 at 7:47 AM Paul Koning via cctalk <
cctalk at classiccmp.org> wrote:

>
>
> > On Nov 28, 2018, at 9:43 AM, Ethan via cctalk <cctalk at classiccmp.org>
> wrote:
> >
> >> As an aside - once upon a time I worked for a company that made their
> own Sparc boards to fit inside a supercomputer and several of them were
> inside secure military/government establishments. Sometimes a board would
> fail and have to go back for a fix - and then the RTC/NVRAM chip had to be
> removed because - you know, those 64 bytes of battery backed RAM might just
> hold some state secret or something...
> >> Fun days.
> >> -Gordon
> >
> > Surprised they knew about it!
>
> One of the documents hardware engineers have to generate is a "Statement
> of Volatility" that lists every component in the system with persistent
> memory of any kind.  For each, it says what is in that memory, where it is
> located, and how (if it can contain anything like user data or
> configuration settings) it can be erased or removed.
>
> The NVRAM chip Gordon mentioned would show up in such an SOV.
>
>         paul
>