Hacker News Reader

Comment by saagarjha | Hacker News Reader

I wonder if you could get the website itself to run on compressed air…

Do you really wonder?

The answer would be "Of course!" and "Of course not!".

You don't need electricity to run a "computer". But as soon as you're trying to hook up this computer to the Internet, serving responses on an electrical network... I can't imagine how to do this without electricity.

Of course you're not looking for answers using a turbine, are you?

For TCP/IP, you can use acoustic modem. It's hard, but not impossible to make mechanical computer and mechanical acoustic modem or vibrational modem (vibrations are easier to parse using mechanics).

You still have to transfer it over the phone line using electricity.

RFC 1149 disagrees.

Here they talk about a miniatirized pneumatic microprocessor using microfluidics:

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2917228/

But: it takes a full 10cm wafer to make 50 D flip-flops(a basic component equivalent to a memory bit), and it works at frequency of 0.5hz .

So it's a challenge.

But maybe it's possible to build a very dumb terminal with this(say keyboard + screen

https://www.nature.com/articles/s41378-018-0018-1

) , if you offload everything to the cloud - and use a really basic communications protocol.

But will the Amish accept a computer/network ? What will it look like ?

Fluidics can work much faster upto 10 KHz[0], maybe even up to 30 KHz or higher if one uses light gases like hydrogen or helium. Although fluidic elements tend to be pretty and have a foot print on the order about a square centimeter. They are also difficult to scale down because it's difficult to maintain the same reynolds number.

Mechanical logic is an interesting option as some newly developed MEMS logic gates might be able to operate at MHz frequencies[1].

[0]https://apps.dtic.mil/dtic/tr/fulltext/u2/a084924.pdf [1]https://www.nature.com/articles/s41467-019-08678-0

Well, fiber optic cable uses light to transmit data... a vibrating mirror could modulate the output. How to receive, well, that seems a bit harder! Also probably some small issues with precise timing and bitrate are to be expected.

Communication systems have methods to compensate for timing errors. And it's probably feasible to build all those circuits at the server side.

But bitrate would be low.

how small these "small issues" will be is of course a question of definition. Details, right? :)

In theory yes, I mean you can make computers using hydraulics - iirc older automatic transmissions use a hydraulic computer.

This one is fascinating:

https://en.wikipedia.org/wiki/Water_integrator

Amazing! I remember reading about a fictional water-powered computer in Terry Pratchett's "Making Money", and assumed it was just used as a fantasy equivalent of electricity; I never realized there were actual computers that worked like this!

it's kinda crazy, but it's fun thinking about how it would affect how you might compute values.

Some things seems like they could become very easy - e.g. addition is just pouring the water from one location into another. Overflow becomes literal overflow...

Subtraction becomes a problem of opening a valve until you've removed the right amount of water... Maybe you could use a weight mechanism that'll close once the removed amount equals the amount you're subtracting?

A precise one and/or a fast one certainly would be very hard to make, and I'm sure the real ones were a lot more complicated, but many simple operations seems like they'd be quite easy to figure out.

Can't imagine it'd be remotely practical at any kind of scale, though... It's amazing they were used as much as they were.

Reply on news.ycombinator.com