I have an idea. Standardize every level/layer of design, from cpu, to
soc board, to computer chassis (not one size fits all, but several for
versatility)

Chip designers are standardizing SoCs using heterogeneous chiplets:
https://semiengineering.com/what-is-ucie/ Possible standards include
bunch of wires (Bunch of Wires), and uci-e.

"Ultimately, Ventana's strongest selling point is bringing the
incremental cost of custom silicon to tens of millions of dollars from
the current hundreds. When hyperscalers only buy some chiplets from
fabless firms and the rest of their silicon directly from the
foundries, the incremental cost per deployed package for a specific
workload is significantly lower.” '"The economics are impressive.
Ventana is also designing its CPU to make it easier to port across
process nodes. "

from https://www.semianalysis.com/p/ventana-risc-v-cpus-beating-next

Combine that with Skywater PDK, and potentially you can have an open
source/libre system of designing pcbs using a standard platform, at
any foundry that supports RISC-V or some other architecture, like
PowerPC (https://libre-soc.org/) One could fundraise a chip that a
sufficient number agrees upon, and potentially lower the cost to
manufacture a chip, perhaps using tools that also allow porting to
lower nodes (i.e 90nm from 130nm, and so on).

Those standards are being designed today, and could very well dominate
the industry for the next 30 years. So I think it's best to see what's
being made to determine what a forever computer's interest would most
closely be aligned with.

> Those standards are being designed today, and could very well dominate
> the industry for the next 30 years. So I think it's best to see what's
> being made to determine what a forever computer's interest would most
> closely be aligned with.

These are really great ideas and lists of current work. If we were to 
look at what it would take today to repair a computer from 1992, we 
would need to be able to produce/procure a motherboard with components 
(ISA cards, memory, cpu). If I could textually show some hand-waving, I 
would do that while saying ISA cards seem to be mostly available, 
creating PCBs for the motherboard and memory seem to be possible with 
generally-available equipment and some off-the-shelf chips.

The CPU seems like the most problematic piece to source reliably for 
procurement. Options I've thought of are:
  - FPGA emulation (like https://github.com/MiSTer-devel/ao486_MiSTer )
  - repurpose portions of the Intel Quark, a 386 soc microcontroller  
(https://en.wikipedia.org/wiki/Intel_Quark )
  - Custom Chip fabrication (presuming we had the specs) 
(https://hackaday.com/2018/04/24/custom-chips-as-a-service/ )

Additionally, a host of digital electronics equipment would be needed to 
inspect and test the hardware.

Some follow-up questions that come to mind:
  - Presuming that fabricating CPU chips is the intended strategy for 
providing maintenance on the 'Forever Computer':
  -- Is it reasonable to expect that fabrication labs would continue to 
exist as a service? (or would this be something that would become highly 
regulated due to 'national security concerns'?)
  -- Is there any benefit to having the ability for individuals to 
own/operate one of these kinds of foundaries?

Also, (forgive me if I'm re-hashing some conversations that I've missed) 
What would an MVP look like for a first step towards a 'forever 
computer'? What's going through my head is a chicken/egg problem, or 
more specifically, software/hardware problem. What comes first?

Thanks for allowing me to participate here :)

Robert

I feel like a computer that can be maintained needs to be macroscopic. A soldering station would be a barrier to entry (I hear this in the ergonomic mechanical keyboard world), but not as high a barrier as a wire bonding machine and scanning electron microscope are.

Unfortunately, part of how we got small, capable smartphones was by abandoning this and other aspects of maintainability. To refocus on this requirement may mean stepping outside the mainstream of state-of-the-art standards. Perhaps I could speak more optimistically of influencing them; but they are so big, in so many ways.

There was a device once, briefly, called the Runcible, which aimed to be upgradeable through time. It also tried to avoid monopolizing your attention, some ten years before iOS devices started giving you screen time reports and Android started trying to ensure your "digital wellbeing." I still want one.

Another aspect of maintainability is the capability to preserve OS support for hardware as OSes change. The FSF's Respects Your Freedom program is an attempt to surface that requirement.

I feel like the forever computer would benefit from being FPGA based
because presumably, treating it like a virtual processor means that
manufacturers would be irrelevant.  SO from a future-proof
perspective, I think this is a good way to go.  It also allows going
'off the map' in terms of how the CPU architecture can run.  It
doesn't have to be derivative of anything existing now.  Sure, at the
moment FPGA's aren't going to out perform an intel chip, and maybe
never will, realistically, but if the goal is for a different
experience altogether, I don't see that necessarily as a problem.  The
only thing it means is cutting edge games might be out of the
question, and folks doing scientific research might have to chain a
couple of forever computers to calculate extreme problems at a rate
that can be done on a stock mass produced CPU.

I mean, if we think about it, an FPGA is kind of an unexplored
universe...the Forever Computer has the potential, going this route,
to pioneering the next 'big thing' by throwing out everything we know
about CPU architecture as we know it, which echos what I mentioned in
another discussion about inventing the future instead of just going
with 'off the shelf solution that exist now'.  And FPGA is one in
which it's both at the same time: off the shelf and has the ability to
be unlike anything else out there.

On Wed, Dec 28, 2022 at 1:21 PM <robert@watkins.net> wrote:
>
>
> > Those standards are being designed today, and could very well dominate
> > the industry for the next 30 years. So I think it's best to see what's
> > being made to determine what a forever computer's interest would most
> > closely be aligned with.
>
> These are really great ideas and lists of current work. If we were to
> look at what it would take today to repair a computer from 1992, we
> would need to be able to produce/procure a motherboard with components
> (ISA cards, memory, cpu). If I could textually show some hand-waving, I
> would do that while saying ISA cards seem to be mostly available,
> creating PCBs for the motherboard and memory seem to be possible with
> generally-available equipment and some off-the-shelf chips.
>
> The CPU seems like the most problematic piece to source reliably for
> procurement. Options I've thought of are:
>   - FPGA emulation (like https://github.com/MiSTer-devel/ao486_MiSTer )
>   - repurpose portions of the Intel Quark, a 386 soc microcontroller
> (https://en.wikipedia.org/wiki/Intel_Quark )
>   - Custom Chip fabrication (presuming we had the specs)
> (https://hackaday.com/2018/04/24/custom-chips-as-a-service/ )
>
> Additionally, a host of digital electronics equipment would be needed to
> inspect and test the hardware.
>
> Some follow-up questions that come to mind:
>   - Presuming that fabricating CPU chips is the intended strategy for
> providing maintenance on the 'Forever Computer':
>   -- Is it reasonable to expect that fabrication labs would continue to
> exist as a service? (or would this be something that would become highly
> regulated due to 'national security concerns'?)
>   -- Is there any benefit to having the ability for individuals to
> own/operate one of these kinds of foundaries?
>
> Also, (forgive me if I'm re-hashing some conversations that I've missed)
> What would an MVP look like for a first step towards a 'forever
> computer'? What's going through my head is a chicken/egg problem, or
> more specifically, software/hardware problem. What comes first?
>

Personally, I think hardware should be looked at early on in a
concrete way in terms of the guts (motherboard, CPU, etc.) because of
the



> Thanks for allowing me to participate here :)
>
> Robert

> >   - repurpose portions of the Intel Quark, a 386 soc microcontroller
> > (https://en.wikipedia.org/wiki/Intel_Quark )

I've always wondered what the Quark could do.
https://ark.intel.com/content/www/us/en/ark/products/86826/intel-quark-microcontroller-d1000.html
(ran between 1-32Mhz and 1.6mW and 25mW)
https://en.wikichip.org/wiki/intel/quark#Microcontrollers
An old version of linux could run on a 386, which the Quark has (I'm
not sure how much cache and RAM would be needed):
https://tiny.slitaz.org/index.php (useful package/iso manager for
building images)
"Useful software, expansible, easy to configure, runs fully in RAM,
simple, light and fast for minimum hardware resources, i.e. fits on
one floppy disk (IDE disk optional), runs on a 386SX processor and
needs as little memory as possible (currently 4MB with a 2.6.14
Kernel)."

Windows 3.1 ran on as little as 1MB of RAM:
https://en.wikipedia.org/wiki/Windows_3.1x#System_requirements (DOS
was in written in Assembly language-not sure about 3.1)
https://www.intel.com/content/www/us/en/foundry/intel-foundry-services.html
Considering Intel opened up their foundry services, do you think
they'd be open to licensing their 386-based Quark? :)


On Thu, Dec 29, 2022 at 9:16 PM Jeremy Landry <hakyoku@gmail.com> wrote:
>
> I feel like the forever computer would benefit from being FPGA based
> because presumably, treating it like a virtual processor means that
> manufacturers would be irrelevant.  SO from a future-proof
> perspective, I think this is a good way to go.  It also allows going
> 'off the map' in terms of how the CPU architecture can run.  It
> doesn't have to be derivative of anything existing now.  Sure, at the
> moment FPGA's aren't going to out perform an intel chip, and maybe
> never will, realistically, but if the goal is for a different
> experience altogether, I don't see that necessarily as a problem.  The
> only thing it means is cutting edge games might be out of the
> question, and folks doing scientific research might have to chain a
> couple of forever computers to calculate extreme problems at a rate
> that can be done on a stock mass produced CPU.
>
> I mean, if we think about it, an FPGA is kind of an unexplored
> universe...the Forever Computer has the potential, going this route,
> to pioneering the next 'big thing' by throwing out everything we know
> about CPU architecture as we know it, which echos what I mentioned in
> another discussion about inventing the future instead of just going
> with 'off the shelf solution that exist now'.  And FPGA is one in
> which it's both at the same time: off the shelf and has the ability to
> be unlike anything else out there.
>
> On Wed, Dec 28, 2022 at 1:21 PM <robert@watkins.net> wrote:
> >
> >
> > > Those standards are being designed today, and could very well dominate
> > > the industry for the next 30 years. So I think it's best to see what's
> > > being made to determine what a forever computer's interest would most
> > > closely be aligned with.
> >
> > These are really great ideas and lists of current work. If we were to
> > look at what it would take today to repair a computer from 1992, we
> > would need to be able to produce/procure a motherboard with components
> > (ISA cards, memory, cpu). If I could textually show some hand-waving, I
> > would do that while saying ISA cards seem to be mostly available,
> > creating PCBs for the motherboard and memory seem to be possible with
> > generally-available equipment and some off-the-shelf chips.
> >
> > The CPU seems like the most problematic piece to source reliably for
> > procurement. Options I've thought of are:
> >   - FPGA emulation (like https://github.com/MiSTer-devel/ao486_MiSTer )
> >   - repurpose portions of the Intel Quark, a 386 soc microcontroller
> > (https://en.wikipedia.org/wiki/Intel_Quark )
> >   - Custom Chip fabrication (presuming we had the specs)
> > (https://hackaday.com/2018/04/24/custom-chips-as-a-service/ )
> >
> > Additionally, a host of digital electronics equipment would be needed to
> > inspect and test the hardware.
> >
> > Some follow-up questions that come to mind:
> >   - Presuming that fabricating CPU chips is the intended strategy for
> > providing maintenance on the 'Forever Computer':
> >   -- Is it reasonable to expect that fabrication labs would continue to
> > exist as a service? (or would this be something that would become highly
> > regulated due to 'national security concerns'?)
> >   -- Is there any benefit to having the ability for individuals to
> > own/operate one of these kinds of foundaries?
> >
> > Also, (forgive me if I'm re-hashing some conversations that I've missed)
> > What would an MVP look like for a first step towards a 'forever
> > computer'? What's going through my head is a chicken/egg problem, or
> > more specifically, software/hardware problem. What comes first?
> >
>
> Personally, I think hardware should be looked at early on in a
> concrete way in terms of the guts (motherboard, CPU, etc.) because of
> the
>
>
>
> > Thanks for allowing me to participate here :)
> >
> > Robert



-- 
Giovanni Lostumbo
708-303-8175