Open Fabs

Zenaan Harkness zen at freedbms.net
Thu Jul 30 16:43:05 PDT 2015


On 7/30/15, oshwm <oshwm at openmailbox.org> wrote:
> On 30/07/15 23:11, The Doctor wrote:
>> On 07/30/2015 11:42 AM, grarpamp wrote:
>> > because you can exhaustively test their logic. On the other hand,
>> > how do you know that once you connect enough of them to each other
>> > that their secret gates inside don't sense each other and activate?
>> > Since you're that close to stone age anyway, why not start one more
>> > step back at relays and core memory.
>>
>> So what you're basically saying is that the entire tech stack, all the
>> way back to far edge pf electromechanical information processing is
>> basically completely untrustworthy.  There is no way at all to trust
>> anything that we can't actually see the logic gates of with the naked
>> eye, which would put us... where?  Maybe tens of computations per
>> second, at most?  A little more (but not much)?
>>
>> Fuck it.  Time to go home, everyone.  They Won.
>>
>
> That is spot on, we can't trust any of it and most people would concede
> that we have lost the battle.

I don't buy that.

0) Whilst using modern Intel vaseline chips:

1) Program full FLOSS stack for circuit/chip dev:
# some starts:
apt-cache search circuit
apt-cache search electron

2) Start with one of the FLOSS CPUs, eg SPARC2, and divide and conquer
it's analysis audit.

3) With open audited/auditable fab, we burn some chips.

4) Now divide and conquer to analyse those physical chips, using
physical analysis one step below the process node - eg 120nm chip,
60nm chip analysis.

As these steps occur, software is developed to facilitate each step of
course. Proprietary software for the audit bits though to make sure it
is not backdoored by Intel.



More information about the cypherpunks mailing list