Kyle Rankin @kyle@librem.one

@aral I try to steer clear of these sorts of things so I don't know much about this particular controversy. That said, all of my (admittedly limited) interactions with Nick from Calyx over the past few years have been positive so at least based on my personal experience something seems off about this.

@ajmartinez That would be fun. At this point I'll need to source a replacement gear at a minimum and then completely remove the control unit to reset all the interlocks. I think the easiest path is to find another parts machine and then make one good machine out of both of them.

@GirthyChode I've been deep into the repair documents and engineering drawings for a week now tracing the inner workings. Fascinating machine. Unfortunately a replacement gear isn't an off the shelf item, so I'm going to have to keep an eye out for another one of these in a similar state so I can make one good one.

Well this is disappointing. While trying to loosen up stuck mechanisms a few teeth broke off of an important gear. Guess this is now a parts machine.

@eliasr @aral @claus In skimming the docs it looks like it lets the user set their own root certs, which is good. LinuxBoot is a reasonable alternative if you have a system that doesn't support Heads.

I still personally prefer the fact that Heads/PureBoot authenticate the host to your USB security token and think that's a bit stronger and more flexible.

I don't dispute that jails are hard to break into, but they are even harder to break out of. How many of us would choose to live in a prison in real life? Instead we make risk assessments that balance personal freedom and security, and the digital world should be no different.

The security industry has largely failed to build security measures without reducing a user's freedom more than attackers. It's not only harder to build security solutions that give users control, it reduces dependence on the vendor. This is why vendors just build jails.

@aral The problem is that the increased security in Apple products come at the cost of control. Security is the excuse but the goal is to limit the customer, not the attacker.

The rest of the industry also wants that control and buys into that security model and considers it the only way to do things.

We take a different approach prioritizing customer control, but it means rejecting many existing solutions and building new ones from scratch. We've made progress but not done yet.

@aral @claus We take a different approach for boot security with PureBoot because I don't believe in the control tradeoff you must make with verified boot and similar "jail vs jailbreak" solutions.

Any solution that depends on blocking binaries the vendor didn't sign anchors too much trust in the vendor and removes control from the user over what software they run.

@rbrown It would literally shock people.

It's pretty complicated all right...

This is a Marchant "Silent Speed" ACR8D electromechanical calculator from the 1940s. It is currently jammed up, stuck on a division problem it never finished, and is apparently the most complicated mechanical calculator out there. Wish me luck!

To close out this thread, here is correspondence between Felt himself (the inventor of the Comptometer) and Zenith Pub. Co.

Even though I have more "advanced" mechanical calculators like my Monroes, the more I use this Comptometer the more impressed I am at the design. It's actually faster to use than my Monroes, at least for addition and multiplication.

Experts would only use the bottom rows so their hand was stationary. To enter larger numbers they'd enter two smaller numbers so to enter an 8 they'd enter 4 two times. This and later models had features to catch "fat finger" errors and partially-depressed rows.

Division was possible via repeated subtraction starting from the left and shifting to the right. There were no underflow bells, you'd just have to know when to stop subtracting a row. This is hard to do one-handed so I didn't film a video of that.

Subtraction was weird. You'd subtract w/ addition by the complements method, but would first subtract one from the number, then press down that number using the small digits on each key. You also slide the correct metal tab at the bottom to stop the carry. Here's 9 - 2.

You can perform many calculations surprisingly quickly and w/ one-hand. A well-trained Comptometer user was probably as fast as you'd get until electromechanical calculators. Here I'm calculating 1024 x 768, which just requires pressing 768 a few times and shifting left.

The Comptometer had a superior design. Pressing a key immediately updated the register, no need to pull a lever (the lever here clears the register). This design was so effective it stayed essentially the same (with minor improvements on error prevention) through the `50s.

This is a Comptometer Model E mechanical calculator, made between 1913 and 1914. Comptometers were made starting around 1900 and were a direct competitor to the 1908 Burroughs adding machine I featured earlier.