Harder Drive: Hard drives we didn't want or need
April 10, 2022 2:40 AM   Subscribe

Harder Drive: Hard drives we didnt want or need - YouTube
In this video we make and evaluate several hard drives that we didn't want. Drawing some inspiration from vexing current events, we find that creative, structured thought on adjacent (but frivolous) problems is a sort of digestive act, and one that is ultimately laxative. Paper and source code: http://tom7.org/harder For SIGBOVIK 2022

How many chainsaws could one feasibly juggle?
Let's build a Hard Drive out of the Internets.
Let's build a Hard Drive out of Tetris.
Let's build a Hard Drive out of used SARS-Cov-2 tests.

Why on earth?
We do these things not because they are HARD, but because they are HARDER DRIVES! - J.F. Sigborvik VII Ph.D.
Lot's of nerd LOL......

Harder Drive: Hard drives we didnt want or need - webpage, paper link, etc.

posted by zengargoyle (29 comments total) 33 users marked this as a favorite
When this popped up in my subscription feed I had forgot why I subscribed... Watched anyways because there was probably a reason.

The closest I could come was being at a very tech conference and building a "giant" distributed RAID drive out of the SWAG USB hubs and drives that vendors were handing out because we're bored at the current presentation. And that I could have pulled off the internet scanning a magnitude faster (shush NDA).
posted by zengargoyle at 3:02 AM on April 10, 2022

yessss new tom7
posted by DoctorFedora at 5:44 AM on April 10, 2022 [5 favorites]

I was anxious that the bits in the covid-test-based drive were going to be positive versus negative results, and that you would write data by infecting people.
posted by fantabulous timewaster at 6:46 AM on April 10, 2022 [11 favorites]

“reliability: low. harm to society: wtf”
posted by fantabulous timewaster at 6:47 AM on April 10, 2022 [3 favorites]

Nice pivot at the end, I hope he convinces a few Blockchain nerds that they are the bad guys here.
posted by subdee at 9:16 AM on April 10, 2022 [4 favorites]

Is there a reason he was soldering stuff while it's plugged in? Twice? Oh wait, yeah, I see it.
posted by kleinsteradikaleminderheit at 9:20 AM on April 10, 2022

Tom7 is a treasure.
posted by vernondalhart at 11:48 AM on April 10, 2022 [1 favorite]

"We can of course ignore air resistance because chainsaws will cut through air like butter."
posted by justsomebodythatyouusedtoknow at 2:28 PM on April 10, 2022 [6 favorites]

Speaking of weird internet things … oh, man, that’s a rabbit-hole … have y’all read about IP over Avian Carriers ? It was based on RFC 1149 and was all fun and games until someone mentioned bird flu. But RFC 2549 and RFC 6214 kinda fixed that issue (in a network engineering sense - we still had bird flu).

Birds had better be real - they may be the future of storage.

Imagine the Easter Eggs!
posted by JustSayNoDawg at 4:45 PM on April 10, 2022 [3 favorites]

"The saws can be efficiently packed in this nice arrangement."
posted by wordless reply at 6:36 PM on April 10, 2022 [6 favorites]

Chainsaws will cut through the air like butter. say this like Howard Cosell.
posted by Oyéah at 8:40 PM on April 10, 2022

Is there a reason that Tom's and SIGBOVIK's sites don't use https? Is this sign of dedication to open source/open internet? "I have nothing to hide?"
posted by bendy at 8:47 PM on April 10, 2022

$ host sigbovik.org
sigbovik.org is an alias for www.club.cc.cmu.edu.
www.club.cc.cmu.edu has address
www.club.cc.cmu.edu has address

Probably has do to with how the Computer Club team has better things to do than to lock down the integrity of a joke conference with a cert system that breaks every 90 days.
posted by pwnguin at 9:25 PM on April 10, 2022 [1 favorite]

The Hofstadter vibe is overwhelming. That's a good thing, mostly.
posted by Ivan Fyodorovich at 10:42 PM on April 10, 2022

What about Tom's long-running site using http?
posted by bendy at 10:43 PM on April 10, 2022

I've been wondering if the chainsaw thought experiment took into account that longitudinally there could be multiple orbits based on the width of the chainsaw and how many could fit into a circle. And still takes advantage of the up/down space that is still available, I mean really the orbiting chainsaws just need a touch to maybe compensate for air friction otherwise once set in motion they'll go on forever all by themselves. Then the problem becomes collision avoidance and/or redundant error correction. I'd bet if you overthink a plate of beans you could juggle a bunch of chainsaws that eventually come back.

Takes a Maxwell's Demon spherical cow (with hands) to keep the information flowing.
posted by zengargoyle at 10:44 PM on April 10, 2022 [1 favorite]

Yeah, it's totally a PITA institutionally wise to get a cert for every host. GRAR war stories.
posted by zengargoyle at 10:48 PM on April 10, 2022 [3 favorites]

I was a little worried that his praise of "the human element" was going to culminate in working out how much data you can tattoo on a person's skin. "More skin means more data, so to maximize data storage, I force-fed each drive in a pit in my basement over the course of several months."
posted by phooky at 11:55 PM on April 10, 2022 [5 favorites]

Welp I've just been laying here awake and thinking about juggling an entire Dyson swarm or constellation of chainsaws. I think we're going to need a second juggler to help refill the gas tanks on all of these damn flying chainsaws.

Unfortunately I think the orbital ephemeris for such a network with multiple intersecting inclination segments of a roughly geodesic satellite chainsaw constellation reaching just above the ground at a single would be utterly impossible due to the rotation and inclination of the Earth.

But I can tell you the noise it's making is well and truly absolutely horrendous. I can barely hear the Doppler effect of the buzzing, flying chainsaws the constant sonic booms.
posted by loquacious at 12:57 AM on April 11, 2022 [3 favorites]

Details. Never mind the world ducking to avoid the flying chainsaws new sky. Now if you could only determine how many times the chainsaws had flipped in their travel... some extra bits of information. Has to be multiple of full flip or you'll end up chopping off your fingers, but there be bits to be had in that rotation of chainsaw dimension.
posted by zengargoyle at 1:52 AM on April 11, 2022 [2 favorites]

Dear authors and editor,

With apologies for the lateness of this review, I recommend that this paper be published after minor revision. It is an unusual and innovative paper. Please consider the following points when revising:

(1) - You never specify, nor reference, the information content of an individual chainsaw. The calculation appears to assume that a chainsaw (or absence of a chainsaw) represents a single bit, which seems incredibly conservative and might lead to impression that it was designed to preference your proposed technology.

The paper would be much stronger even if a naive model that incorporates conservative machining tolerances and chainsaw tooth length were included.

The practical difficulties of operating chainsaws in vacuum also should be discussed. Assuming electric motor chainsaws, overheating should be addressed.

(2) - In the section 1.2, the assertion of 40Mb/s appears to assume the entire 0-40 MHz band will be available for use at all times, with some questionable assumptions regarding encoding and error correction. This seems quite optimistic and would be challenging to implement on Earth for political and historical reasons. A more realistic model, or set of models, that is compatible with existing digital protocols would be helpful to the reader who intends to continue this research. Furthermore, even assuming sole access to the entire 40 MHz band, only a small fraction of the band will experience ionospheric bounce at a single time of day. For future work, I would suggest considering the possibility of active ionospheric modification to improve this.

Also, while it may be outside of the scope of this paper, please consider multiplexing through the use of directional antennas in the future. While it is unlikely to buy you the factor of tens of thousands needed to justify the assumptions discussed above, it can't hurt. You might also consider moon-bounce options in future papers.

(3) - While I would not require this for publication, it might be worth considering passive slowing techniques when approaching these systems implemented at scale. Switching from fibre to a high dielectric constant (and or high-mu, ferromagnetic loaded) transmission line could improve performance by a factor of several beyond these impressive results.

(4) The Tetris implementation is beautiful work and the authors should consider making it a stand-alone paper.

(5) The CU experiment appears to be little more than an implementation of an already-existing and commercially manufactured storage device. Perhaps this would be more appropriate for a proceedings or a more applied journal.

[In case it's no obvious to other mefites, I love this. Thanks for posting it!]
posted by eotvos at 7:07 AM on April 11, 2022 [9 favorites]

I could have sworn I saw a TAS Tetrisprinter that would probably have saved him some time on the Tetris Drive portion, but I'm not seeing it around any more.
posted by Kyol at 7:38 AM on April 11, 2022

Also re: the Cue cartridge drive (so near and dear to my heart), while this is most certainly a Harder Drive, I feel they could have gone a little further down the road to make a Hardest Drive by designing a single socket that cartridges could be reliably inserted and removed from manually (or not so reliably! Maybe you'd have to blow it out with compressed air once in a while). Not only would it have been more flexible (and infinitely scalable!), it would have required less soldering, reduced read and write speeds (assume an insertion/removal time of 0.75s), and required a bit more of that human touch.
posted by phooky at 12:21 PM on April 11, 2022 [2 favorites]

I love that the whole thing turned out to be a setup to dunking on cryptocurrency. Great job tom7, no notes. Okay, one note, please get that basement outlet that's trying to kill you fixed.
posted by biogeo at 1:44 PM on April 11, 2022 [3 favorites]

Question for folks here:

Around 6:40 in the video, Tom7 talks about a block of addresses that didn't respond to his pings because they were for IP multicast (IPV4 to and then he outlines the block and writes 'dark web' w/o any further explanation. Perhaps someone here can clarify this because it doesn't jibe with my understanding of the workings of the actual 'Dark Web' and onion routing. Was he just being funny?
posted by Insert Clever Name Here at 3:24 PM on April 11, 2022

Yeah it was just a joke about that section of IPs all showing up in black because they didn’t respond
posted by DoctorFedora at 3:30 PM on April 11, 2022 [5 favorites]

My friend says, "i want the custom LUA scripts this guy must have written for this bit so bad," re: the tetrix parts. Suggests a LUA script for the NES and information about how to compile it.
posted by subdee at 5:22 PM on April 12, 2022

Late to the party but this is delightful. (A friend mentioned it to me in passing because of the Hilbert curve storage of the IP map, heh.)
posted by cortex at 10:30 AM on April 27, 2022

We did the Hilbert curve quite a bit. I had two /16 networks that I could pop up a 512x512 or 1024x1024 image and just look for things not being 'right' somewhere and catch things before they fail enough to get a call but before they trigger alarms. Look for patterns, reshuffle by switch and port, "ah, that switch is starting to fail" or "janitor or something bonked those connections somebody needs to take a look". Look into the matrix, there's something wrong in this picture. The Hilbert sort of locality curve makes it easier to spot before it gets bad. I'm still surprised that everybody doesn't do that, way too many W x Z grids that break the locality principle..
posted by zengargoyle at 1:46 PM on April 27, 2022 [1 favorite]

« Older I want to marry a lighthouse keeper. Just not that...   |   “feel their grief & also take action that’s... Newer »

This thread has been archived and is closed to new comments