The Computer at the Bottom of a Canal | Hacker News Reader

The Computer at the Bottom of a Canal

negroniventurestudios.com

183 pointsKudos48 comments

Very interesting and well written article. This sentence in particular:

"Linn Products is the Glasgow company Ivor Tiefenbrun founded in 1972, and if you know it at all you know it for the Sondek LP12, still widely regarded by its partisans as the finest record deck ever made."

I was a big audiophile right at that time. Auditioned both Sondek and Sota Star Saphire. Went with the latter which I still proudly own (Alphason Tonearm). Note how carefully the author phrased "...by its partisans..." when describing the Sondek as the best. Nothing, not even programing code style zealots today can compare to audiophiles and their "take" on what made for the best playback. Pretty sure the author did not want to open that can of worms by simply declaring the Sondek the best.

Turntables back then were pretty damn detailed re: technology. Not surprised Linn could make a great computer (in the day).

HTH, NSC

Yeah but also saying "it's considered the best by the people who love it" is a tautology.

Therein lies the humor, imo.

Thing is capability machines (like this and the 432 and lots of research machines) were very much the thing at the time - cutting edge even. The research literature was full of them. I did a paper design at the time.

What ate them up was "what can you fit all on a chip with not many pins", followed by "what can you fit along with a cache on a chip with more pins", things move so much faster if everything's on the same die.

Tagged architectures are old, Burroughs mainframes had them back in the 70s along with rudimentary hardware objects (pageable even)

The end of the article seems to admit all of what you say but then suggest it is time to revisit; do you think that’s valid or no?

I'm in 2 minds, I'm strongly in the RISC camp these days, I know how hard it is to build a truly performant CPU and think that simpler is better as far as ISAs go. Adding segmentation hardware into the addressing path means more clocks and makes it's harder to detect loads/stores that interfere with each other (all load/stores have to block each other until the addressing/tlb lookup has been done ie detecting aliasing ).

A true capability architecture needs separate register files for pointers and data with protections enforced when the address registers are loaded/stored, it's a step away from our models of linear address spaces that many of the languages we use today are based on. And any new CPU that requires everyone to learn a new language is probably a non-starter as a mainstream CPU

If only the LLMs have to learn the new language then maybe it can be a starter?

They are old and still around, even though communities like HN tend to ignore them.

IBM i, z/OS, ClearPath MCP, and ClearPath OS 2200 are still out there, getting new releases, every now and then.

SPARC ADI is still being sold by Oracle and Fujitsu.

ARM MTE and Pluton are shipping, even if in not all models.

Can you expand on your paper computer? It sounds fascinating.

It was more than 40 years ago, before I had the skills to build chips, I'm afraid I've kept nothing of that design

No worries. Before your comment, I didn’t even know about this type of computer. This is my new rabbit hole. Thank you for that.

It’s crazy what a small group of smart people can achieve if they’re not captive to conventional wisdom.

They got crushed by the commodity curve and Moore’s Law. But still a great story - and thankfully written by a real person.

The author’s idea that the commodity curve is over and that hardware is now cheap enough to make special purpose hardware viable is intriguing. Standardization used to be important because you needed to convince scarce programmers to invest in the platform, but AI has ended the scarcity of programmers so dedicated purpose hardware is more viable than ever.

There are several cycles that bring tech ideas in and out of fashion because of market conditions, intersecting tech cycles, other implications of Moore's Law, or whatever.

Another trend, for example, might be centralized vs distributed: terminals (thin clients) and mainframes gave way to fat client PCs on prem, and then back to thin clients with cloud stuff.

I wonder if you could throw a small microcontroller at the bottom of a canal, powered by water passing through a fan, to hide it. And why you would want to do that.

canals have negligible water flow (except during disasters!) so I don't think that'd work very well. They also get dredged occasionally so there's a decent chance anything you leave down there won't last more than a couple of years

> They also get dredged occasionally

And it's fascinating to watch!

https://www.youtube.com/watch?v=aGFC6MF5-_k

They do have tides though, so make it a big fan that points up.

Depends on the canal. The UK narrow boat network has locked everywhere so minimal current in most sections similarly the Netherlands. In fact most canals. But then you have something like Suez which is just open to the ocean.

The UK locks themselves are water movement though, so if you wanted to power something for "free" on those, you'd attach it to there. Oh man did I love watching those as a kid. You wouldn't do a fan though, you'd have a line and a buoy and use the up and down motion to drive a linear electromagnetic generator to harvest energy.

Yeah but it's not really a lateral flow of water when the locks fill up and empty.

you do get lateral flow up/downstream during filling and emptying, but it's generally considered rude to do it fast enough to cause turbulence

Water quality monitoring. Communicate back to the surface with ultrasonic pulses.

And in case people think ultrasonic signaling isn’t a real thing, it very much is. In the oil and gas industry, it’s how measurements taken while drilling a hole are sent back to the surface through the drilling mud.

Downhole tools used in MWD/LWD do not use ultrasonic frequencies. Drilling is a very noisy environment. The first tool that I used employed a downhole pulser with a surface transducer and operated at about 0.5-1 Hz. That's not ultrasonic. It was a pretty low frequency signal. Data transmission rates limit the volume of information that can be transmitted from the downhole tool to the surface where they are decoded.

Your tool has to set up a pressure wave in the mud column that is detectable miles away from the tool while the well is being drilled.

This is a good recent paper authored by an experienced drilling engineer about MWD/LWD drilling history and pulser designs and data rates that can be expected.

https://www.aade.org/application/files/1917/4604/2319/AADE-2...

Oh derp. It’s been a while since I worked adjacent to this, so I clearly misremembered some important bits. Thanks for the correction!

Boy how I wish it had used ultrasonic frequencies. Sadly enough though, I spent large parts of every day in front of a laptop connected to a transducer waiting for signals to appear in the noisy stream of data coming from the standpipe transducer. Every few seconds if your tool still functioned you would get the first bit of a signal that contained toolface orientation information and gamma ray data if you had a gamma ray sub. It was a binary stream of ones and zeroes where a 1 was a pressure increase and a zero was a decrease to background. You needed to be quick on the decode because the directional driller on the rig floor was using the real-time toolface orientation to help steer the drill bit to the target formation and on to TD (total depth).

As the days went by and the hole was progessively deeper it was inevitable that some part of the downhole BHA (bottom-hole assembly) would fail due to erosion from the drilling mud, erosion at the bit/formation interface, friction along the borehole wall, etc. The hope was always that your tool would function well enough until something else failed and necessitated a round trip to replace a critical component. The costs of the round trip were placed on the service company that had the failed equipment so even if your company supplied the MWD/LWD services and the directional drilling services (DD), if your tool failed the cost of the trip fell on your service and impacted your service quality assessment which affected your rig bonus calculation.

It was common for a DD with a failing mud motor to hang around the MWD shack or keep the MWD engineers on the phone, harassing the engineers to get them to admit that their tool is nearing the point where the toolface data was so low level that it was undecodable. If he could hang the trip on the MWD he would every time so the MWD engineers were incentivized to struggle through poor signal for as long as possible when they knew from the pressure data that the mud motor was about to crater or that the drill bit needed to be replaced.

Battery failures and erosion of the pulser valve components that resulted in signal amplitudes diminishing to the point where they were undetectable were the things you knew would happen and you just hoped that downhole conditions killed someone else's tool before yours inevitably failed.

Damn, that's fascinating. What do you think of Fervo energy and their fiber-optic sensing?

Thanks for reminding me that these guys exist.

I had to read a little since I had not checked on their progress in a while.

First, I like the backgrounds and experience levels of the key personnel in the company. Many are oil and gas industry veterans with a good mix of skills - drilling, deviated well design, fracking, downhole monitoring sensors, etc.

DAS deployed downhole as their paper describes should give excellent information about conditions near the boreholes as they pressurize and depressurize the formations through the enhanced fracture networks. Induced seismicity is a common side effect of fluid injection and anyone who has watched USGS earthquake data for California has likely identified the geothermal fields by their frequent seismicity. Reducing the levels of induced seismicity related to geothermal energy production will decrease the likelihood of formation damage due to accumulated stresses in the production field. It sounds like they are able to tie specific stress/relaxation events to pressurization events so that is a very useful tool.

A lot of the earthquakes in the Permian Basin area are related to fluid injection. A DAS network could possibly mitigate some of that so a technology transfer back to the O&G industry may benefit all of us.

I agree with their goal of expanding geothermal energy production as an integral part of the power generation grid since, like they say, it is completely renewable. It does take advantage of downhole conditions in order to produce the energy and there is an opportunity to utilize existing boreholes as geothermal generation sites after the oil and gas becomes uneconomical so it could give new life to old fields and preserve jobs in the area since there will already be a trained workforce locally.

Pretty interesting. I lost track of these guys and now I see that they IPO'd a couple months ago. I expect that their business and their competition will both be increasing over the next few years at least.

Website design comment: I love the orange/teal theme! Really drawn to the details such as illustrations carrying the orange/teal theme. Inspired me to re-color a project with this split complimentary color.

I personally hate the colors; I don't know if it was vibe coded, but it resembles the browns that GPT loves a lot.

I'm looking at it from a crappy screen, though.

Ok, I forgot about it, that was with dark mode enabled. Without it it's much better.

I read about them in Byte Magazine - the names were memorable - Objek - Numerik and a third chip - there were no silicon afaik but jsut emulator circuit boards run by a Sun workstation.

Objekt, Numerik, Logik, Klock, according to Wikipedia [1]

[1] https://en.wikipedia.org/wiki/Rekursiv#Physical_packaging

I console myself that this is why I never got fabulously wealthy. That brilliance is no guarantee of wealth. You need luck, too.

But the truth is, you also need to do the work. I used my conclusion about luck as an excuse to not even try. Not even in some grand way. Just in the grinding everyday way. In the way of, "You know what I should do?! I should ... meh. That'll never work."

Take it from me, kids. If you want to do something great, do it while you are still young enough to believe it's possible.

>In my essay a distribution of one I argued that bespoke software was the original, correct arrangement, that fifty years of productised general-purpose software were a compromise forced by the economics of scarce programmers, and that AI has ended the compromise.

I would also say that productised general-purpose software was an undue bonanza taking unfair advantage of copyright, which shouldn't have allowed so many restrictions on code which is necessary to make devices perform their intended tasks.

Fortunately AI has been able to get started accomplishing some of the much-needed workarounds to these annoying copyright issues, like few humans have been able to do.

Regardless of whether programmers are scarce or abundant.

On a tangent, it sounds like a famous canal, but I would figure there are a number of little-known waterways where there might be a high-performance 21st century PC resting underwater along with some poor soul's bitcoin wallet :(

The "A Distribution of One" argument seems less convincing than the "The Computer at the Bottom of a Canal" one. I'm no expert, but if you've ever read some of the computing press from the time you'll know that custom software was still rather popular well into the '80s: and unfortunately custom software written for in-house use by a non-software company was the classic venue for all the Software Crisis war stories, which probably has a lot more to do with why it went out of fashion than progammer wages on their own. Whether the main problem was the difficulty of extracting requirements from users and management or the lack of sufficiently good software developers, maybe the arrival of LLM software cannons will be what it takes to fix it, but a hae ma doots.

What remains of this computer? Diagrams? Code? Anything that would let us take a closer look at how it works?

The 40-bit hashed address seemed familiar, then I realized that I (and Claude) had used a similar concept in a VM we built for the OS I've been building. It's a capability and typing system that enables import boundaries to be enforced and third-party packages are unable to perform security sensative operations without being specially signed. It's also a "distribution of one" concept as the user builds their own UI to their specification personalising the OS.

A fascinating story.

I owned a Linn Sondek turntabe in the 70's with a Denon cartridge.

Spent a crazy amount on Hi-Fi chasing the audiophile dream :-)

I don't think this is AI, but the writing style keeps reminding me of AI. I'm trying to isolate why. I think it's something to do with the information density. The sentences twist and turn, have lots of related ideas, digressions, and little details without much lead-in. The wording is also very precise, which is another characteristic of AI writing. This sounds like a critique, but I actually find it charming.

This sentence is a strong illustration of what I'm talking about:

> Since only OBJEKT ever knew where anything physically lived, objects could be relocated freely without touching a reference, so garbage collection went into the silicon too: a two-space compacting collector that walked the live objects and slid them into the other half of DRAM while execution carried on above, oblivious.

I mean, come on. Run-ons are one thing, but this is practically a lecture in a single sentence. :P

Very well-written text, not a whiff of AI generation to it.

Your example is not even a run-on sentence!

I enjoyed reading it, many thanks to the author.

I really enjoyed it, but I think there’s more than a whiff.

> That curve is gone. The compromise is going with it; and the workload-shaped machine, the thing Harland was building in 1984, is what comes back.

Nope, and Pangram agrees with me, 100% human written with high confidence.

I really enjoy this style and I think it's not _that_ close to AI output (it's much better in fact).

I was never arguing that it was AI-written.

Right, this was in response to tclancy.

Yeah I'm not sure, just identifying that subjectively, this is the impression I get. Maybe I just associate any style which is especially fluent and avoids vernacular with AI at this point.

Too bad british goverments had to kill companies like that. But the new business plan is going much better!

Your embarrassment could have been avoided by reading TFA.

Reply on news.ycombinator.com