Movies aren't reality, in movies stuff happens because it forwards the plot, not because it is practical or sensible.

In the real world stopping people, especially innocent bystanders, getting killed or seriously hurt normally* takes precedence over almost everything else. If you build a robot that gets broken you lose at most a robot. If your robot kills someone then at the very least you will be facing and probing, distracting and expensive investigation from regulators and at worst could be facing major fines or even prison time.

So building a heavy and rigid robot that protects the robot at the expense of the person it runs into really isn't an option for a robot that will have to share space with the generic public.

* Yes there are exceptional situations, and there are also things that are tolerated for historical reasons that almost certainly wouldn't be tolerated if they were newly introduced today.

The question is will the revenue from such customers justify the costs (both in terms of equipment and spectrum) of keeping at least one 2G network online.

Apologies I misread your post as "It'd" rather than "I'd".

That doesn't follow. It's both perfectly possible and technically reasonable to design a service that does the initial retrieval on demand, but then keeps it's own copy for further use.

Wikipedia claims that sci-hub went through three phases, in the first phase they only did real-time retrivial, in the second phase they stored copies on a third party site and nowadays they store copies on their own server.

Whether wikipedia is correct, I cannot say.

BGP4 was introduced in 1994.

https://tools.ietf.org/html/rf...

$100 per month for unlimited traffic is very good by satellite internet standards. It is however expensive compared to what most urban users pay for broadband.

As you say it's a matter of speculation what their long term pricing strategy will be, I suspect a lot depends on whether they can drive down the cost of "dishy".

IMO cython is more "python with C/C++ bolted on" than "statically typed python". It essentially has two completely seperate type systems, a static one from C/C++ and a dynamic one from python, afaict it doesn't provide a mechanism for using python types in a static manner.

It's hard because even the fastest EV chargers have a much lower effective power than a gasoline pump.

A gasoline pump in the US delivers about 38 liters per minuite and gasoline has an energy density of about 34.6 MJ per litre. That works out to 21 megawatts. Now EVs are far more efficient than ICE vehicles, https://www.fueleconomy.gov/fe... claims 77% for EVs and 12-30% for ICEs, lets take the middle of that range and say that EVs are about 3.6 times as efficient as ICE vehicles so a gasoline pump is equivilent to roughly a 5.8 megawatt EV charger.

The fastest EV and charger combinations are a few hundred kilowatts but for a mixture of different EV types an average of 100KW is probablly more realistic, so about 58 times slower than a gas pump. That means two things.

1. Even taking account of the much higher efficiency of EVs than ICE vehicles, you need a lot more chargers than gas pumps to deliver the same amount of useful power to customers.
2. Charging takes a sufficiently long time that people will be reluctant to just sit around and wait.

Put these together and it makes far more sense to deploy EV chargers in the carparks of supermarkets, shopping centers, tourist attractions or parks than it does to deploy them at gas stations.

I don't know what the rest of Europe is like but here in the UK we have two different connectors for DC charging, (CCS and Chademo), plus some vehicles that are AC charging only. So rapid charge stations end up with three separate connectors to cover all the EVs.

The starlink dish aims itself, it pretty much has to because starlink is not geostationary so sattelite locations are constantly changing.

As I understand it current starlink beta services are registed to a single location and are likely to stop working if you move the dish too far from that location because the sattelites won't allocate you time slots correctly. Exactly what too far is doesn't seem to be clear.

https://www.nsf.gov/statistics... says about 2/5ths of those getting STEM PhDs in the US are citizens and permanent residents. So less than half but certainly not "a small minority".

And then there are ITAR restrictions which make it very difficult for US aerospace companies to hire people who aren't US citizens or permanent residents. https://qz.com/794101/elon-mus...

You don't remember correctly. The first successful falcon 9 landing was at cape canaveral.

Because in many places they are the least bad option. Especially if you are focused on download rather than upload speeds.

Realistically in most places (and not just in the US) your choices for affordable last mile communication services are the incumbent telco* and (if there is one) the incumbent cableco. It's often very difficult for new providers to establish themselves both for regulatory reasons and simply because many of the costs of running a network scale more with the area served than than number of customers.

Cablecos have generally been far better able to scale download speeds on their existing infrastructure than telcos have. So in many places the speeds available from the cableco are significantly faster than those available** from the telco.

* In some parts of the world telcos are forced to work with multiple ISPs, I've never heard of the same being required of cablecos
** At a price acceptable to home/small buisiness users. Obviously if you pay enough money most telcos will drop in a dedicated fiber line.

At least in the UK they say they will ban sales of new fossil fuel only cars in 2030 and plug in hybrids in 2035, of course that multiple general elections away so whether it will actually happen is an open question.

But they haven't said anything about existing cars, It would be a disaster for carmakers if the general public decided the way they were going to deal with an unpopular mandate for new cars to be all electric was to get their existing cars fixed instead of buying new ones.

They had high-end models sure, but they also had much cheaper stuff. Perhaps comparable to Samsung today.

And it's not just Nokia. Afaict of the major mobile phone vendors in 2005 ( https://www.networkworld.com/a... ) only Samsung is still a major player today.

According to wikipedia the 8086 began design in 1976 and was released in 1978. The arm1 was produced in 1985 but afaict was only used in internal acorn development hardware. The first arm chip to make it out into the wild was the arm2 which was apparently created in 1986, so eight years after the 8086.

Arm has also been more able to shed cruft than x86 has. On arm "26 bit" mode* and FPA are long gone, 32-bit mode is still in most current arm chips, but it's days seem to be numbered. 64-bit mode is a fairly radical redesign of the instruction set.

On x86, the x87 FPU is still very much present, and the 32-bit and 64-bit instruction sets are fairly direct extensions of the original 16 bit one, which is also still present.

* Arm always had a 32-bit datapath, but originally addresses were only 26 bit with the high bits of the program counter used for status flags.