nas

I think the "by Linus" is a bit much. He was the founder of the project but others did a lot too. Dirk Hohndel has been the project lead for a long time already.

> The most obvious difference is that Git's index is a bit boneheaded, so I'm ignoring it.

I use the git index nearly every day. I suppose it is possible to implement something like git without it but I would find it quite a bit less useful.

Gee, I wonder if this tweet is in any way connected to his previous one about weakening the USD to match the (alleged) currency manipulations of the Euro and China?

Please hang on to those US dollars. They are very solid, strong and stable. If the S&P 500 hits all time highs, its certainly because everyone is getting so rich, and not because we are inflating the currency. ;-P

Ken, I love watching you and Mike do your stuff (Marc too but usually he is behind the camera). Thank you for taking the time to explain things.

All those guys are amazing but I especially enjoy Ken and Mike reverse engineering stuff. Really impressive.

Valve's revenue models are going to kill it. At least, kill it as a studio that can create high quality games. I not opposed to them making money and getting compensated for hard work but the level of greed is getting insane. For the Dota TI prize pool, they take 75% of the revenue from Battle Pass related sales (i.e. digital in-game items, like DLCs). Its not hard to understand why they do it, the profit is insane:

https://dotesports.com/dota-2/news/the-international-2019-pr...

However, why should the company ever spend the effort to build games? Making the Battle Pass content is vastly easier. I don't see how they are going to be able to wean themselves off such a juicy revenue stream.

Artifact was a result of this new reality. The game was designed from the get-go to optimize revenue from in-game digital items. Why should a digital card in a computer game cost $100? A rational person would realize it is an insane idea. I'm glad that Artifact crashed and burned.

I don't think Valve learned their lesson. The whole game industry is moving that way. E.g. free to play games with paid in-game items. I wish they could figure out a more healthy way to fund game development. Human nature being what it is, I'm not holding my breath.

This is a really nice video if you are interested in measurement precision for manufacturing:

Machine Thinking: Origins of Precision and first project introduction https://youtu.be/gNRnrn5DE58

Workflowy could do this.

Yeah, Curious Droid is a nice channel. I really enjoyed this video on SLAM: https://www.youtube.com/watch?v=DZHONQAMV48

What an insane idea that was.

I hate the term "mining" for that reason. It is not like mining gold. I would have preferred if they called it "digital notary service". That's what the miners are doing. They are notarizing transactions so people can agree on the ledger and so that double-spends cannot happen. Early in the life of the system, the notary service is subsidized by inflation. Later (after many halving of rewards), they will be compensated with transaction fees.

If someone can devise a way to perform the notary service without proof-of-work, I'm pretty sure Bitcoin will move to it. Right now, no other system has been shown to actually work. Proof-of-stake systems are in the works but, as far as I know, none of them are considered trustworthy at this point.

Edit: I would like to also say that I'm sympathetic to the opinion that Bitcoin mining is too wasteful of resources (i.e. electricity). I think the system is perhaps flawed in terms of the connection between Bitcoin price and the economic drivers of difficulty. When the price is high, the mining rewards are such that there is huge pressure to increase mining costs. Do we actually need that level of security? OTOH, hard to blame Satoshi for the design since it is hard to foresee how quickly the system gets adopted.

The beta is now out. The big incompatible change is probably PyCode_New() (in order to support positional only parameters). At least Cython is broken by this change, probably other extensions. Hopefully it will not be long for that to be fixed.

Lots of other neat changes. The f-string debugging with '=' will be handy. There are a number of minor speedups and so this should be the fastest Python yet released.

I think a solution is to create a gradient of radioactivity, so it gets more and more dangerous as you dig deeper.

Some comments based on my understanding of the situation. I'm an engineer and I'm interested in the study of engineering failures. So, I have been following this. Initially I was thinking the blame was being unfairly placed on the MCAS system. However, now I understand more, I suspect it is a serious safety problem.

The reason is as you say, MCAS runaway (e.g. due to a failed AoA sensor) does not behave like a normal runaway trim (e.g. continuous trim movement). That is key. Watching Youtube videos was helpful in understanding what the pilot experiences. I initially thought the pilot should easily notice the trim moving because of the large trim wheel beside them. However, in normal takeoff, the trim is being adjusted automatically and so pilots learn to ignore it changing. Second, I think with the 737 MAX, the sound made by the trim wheel has been changed (removed)?

Using the stab trim cutout switches is a drastic measure (occurs as nearly last item on memory checklist) and, as I understand it, the only way to stop a misbehaving MCAS from doing its bad work. If the pilots are even a little slow in figuring out what's happening, I can easily imagine that leading to disaster. Also, if they to realize the problem and flick the switches off, manually cranking the trim wheel the other way takes time and it takes some serious physical strength to do it. I can imagine that if the MCAS moves the trim fully the wrong way, the pilots might not be able to fix it quickly enough.

This Seattle Times article is quite informative:

https://www.seattletimes.com/business/boeing-aerospace/faile...

It seems the MCAS had a flawed design where it would move the trim further than expected and would also trigger multiple times. Based on my understanding, the MCAS would have been quite safe if it didn't have these flaws (pilots deal with a failed AoA sensor as another case of "runaway trim").

Yup. The question is not if computer automation can be made 100% reliable. Instead, can it be reliable enough such that there are fewer accidents vs manual controls? The answer seems to be yes. So the trend is to more computer automation, not less. I think people who think we can go back to fully manual controls are living in a fantasy.

There is a large trim wheel (beside the pilot, about 10" in diameter) that the pilot can see turning if the system is automatically trimming. It is annoying how much speculation and misunderstanding is going on regarding that system. It might be a factor in the two crashed planes but we don't that yet.

Is he talking about the trim wheel? I don't know much about airplanes but that seems possible to me. Would the trim wheel get the spring force feedback like the yoke?

Edit: here is a video of the trim wheel on a 737-300 https://vimeo.com/34501723

MCAS is not a fly-by-wire system. Pages and pages of speculation by arm-chair aircraft designers going on. So depressing.

So much speculation is going on. The angry mob has already concluded that the MCAS must be the cause. Greedy Boeing was too cheap to redesign the plane and made a cheap software hack to fix it. Sneaky computers are overriding the decisions of the poor helpless pilots. The FAA is too incompetent to certify aircraft, etc etc.

Based on what I read, the truth is a lot more complicated. The MCAS doesn't work the way most people seem to think it does. Maybe it is a factor in the crashes. We don't know that yet.

Your explanation doesn't make sense to me. How many watts does a non-LED light consume vs charging an EV? A good sized light is probably less than 500 watts. To charge an EV you need thousands of watts. The street lights must have been wired with a massively over-designed electrical conductor (and upstream infrastructure) or there would be no way to charge EVs in any kind of density. Based on the power draw of EVs, I don't see how it can work without putting in new cabling.

After years of having AMD video cards, I recently got a high end Nvidia card, GeForce GTX 1070 Ti. For 3-D gaming, Nvidia cards were clearly better performance vs price. However, dealing with the proprietary driver is just a pain in the ass. With an AMD card, the open source driver just worked. With Nvidia, I have to keep re-installing the proprietary driver. Sometimes it doesn't compile with the current kernel (e.g. Debian testing). Finding the correct driver to download and install is a pain (try using the 'links' on nvidia's site).

If I build another PC, I'm going to very strongly prefer an AMD card.

If your Python 2 code is clean and your Python 3 is ugly, I don't know what the hell you are doing. They are almost the same language. Are you trolling?

Analogies are fun. If you have a tower of sand that collapses, do you find the grain that started moving first and put the entire blame of collapse on it? Your example is of relatively stable system (presumably healthy human life) that takes a big external disturbance to cause failure. The sand tower is an unstable system that takes a tiny disturbance to trigger failure. I think the California forest situation is somewhere between these two extremes.

Yeah, I think that's his idea. So, instead of tables of molar mass in g/mol, you would have tables in terms of "amount of substance"/g. E.g. number of atoms or number of molecules.

Yeah, I agree it is not an open source license based on rule 9 alone. I mean, the reason for the license not being open source is not an accident. It's not some unforeseen side effect. Their whole purpose of changing the license is to restrict the freedoms of the service companies using the software. I'm somewhat sympathetic to MongoDB for wanting to extract money from them or to force them to work with the open source community. However, you can't have your cake and eat it too. Forcing them that means your license is not open source.

This has been my experience as well. 20 years ago, you could go to Walmart and buy good quality brands at good prices. E.g. you buy a Dewalt drill, it is actually a high quality drill. What happened over time is that because of the extreme price pressure Walmart was putting on suppliers, they began to build special Walmart-only products. So, the Dewalt drill you buy has a great price but is also a piece of crap. I don't enjoy shopping at Walmart at all anymore, you are much better off to find a specially store. E.g. buy your drill the same place the professional carpenters shop.

Amazon has seen a similar decline. Years ago, if you saw a product on Amazon, you could be confident it was as it was labelled, you would get it in a reasonable time and the price was about as good as buying anywhere else. Amazon quality now is total garbage. The store is littered with counterfit goods. If the price looks good, that's probably because the quality is absolutely terrible. I don't enjoy shopping there at all anymore.

I enjoyed this video a lot (how to eat it properly):

https://www.youtube.com/watch?v=kDr0Kbbw6D0

I love his voice and his accent. I've never tried it myself but I would be willing to if I could buy some. Eating it straight out of the can seems like eating horse radish straight from the jar. Of course it is going to taste awful.

Slackware and 3.5 inch disks for me too. If I remember correctly, I was going to install SLS Linux but this "Slackware" thing was the new hotness. Time flies having fun I guess. Still running Linux on my workstation.

Guido's glasses are much cooler, IMHO.

I'm saying it happened and Python 3 is the language I want to use. If you don't do Python 3, you either keep the error prone str/unicode string model or you dream up some way to evolve to a better model. Would some mythical version of an evolved Python 2.x be better than what we have? I mean, how to argue against that? I haven't seen anyone propose a workable migration plan. Python 3.7 is better than 2.7.15 and better than 'tauthon'.