If you have a recent version of the cpuid tool, you can run:

cpuid |grep RTM

and you'll see something like:

RTM: restricted transactional memory = false
RTM: restricted transactional memory = false
RTM: restricted transactional memory = false
RTM: restricted transactional memory = false

/proc/cpuinfo doesn't show it, presumably because no kernel support is needed at all for this feature. (And that's why, if this is indeed a privilege escalation issue, it won't be easily fixed with a kernel change.)

This seems unlikely to me. I'd expect that the patch will clear the cpuid bit for TSX and cause #UD (undefined opcode) on XBEGIN, etc.

Buying an R7000 is a waste of money. The whole "AC1900" class is a joke -- they squeeze out the extra Mbps by supporting a nonstandard, Broadcom-only modulation in the 2.4 GHz band. Good luck finding any client devices that can use that modulation.

Stick with AC1750 and stop feeding the marketing trolls.

The TP-Link Archer C7 v2 is pretty mature. The v1 is discontinued, and most US-based vendors are now explicitly advertising v2 hardware. The only real caveat is that the open-source ath10k support is very slightly flaky -- you may have intermittent issues if you have a Macbook Pro. This will probabaly be fixed soon.

Do you consider the TPM acceptable? I have sort-of-working but woefully incomplete code for this. There's also the work-in-progress OpenCryptoChip.

This is incorrect. You can't build a passive device that is 100% transmissive from one side and 100% reflective on the other side.

You can, however (in theory), build a device that with these properties:

1. Light from side A is transmitted and comes out side B
2. Light from side B is absorbed and turns into heat
3. Absorbed heat is re-radiated, mostly from side A

This thing will be non-reciprocal. I've never heard of anyone building one that looks like a sheet of glass, but they're common for microwaves -- they're called circulators. There's a reason that this new contraption is called an acoustic circulator. The only misleading part is that one-way glass is reciprocal -- this new thing is much more impressive than one-way glass.

You can do things like this (with a little bit more care) to generate numbers that can't be predicted in advance. Unfortunately, that's not the point. Web servers need random numbers that can't be guessed or manipulated by anyone, and they need to generate lots of them. If everyone generated the same random numbers (because they looked at the same webcams), then those numbers aren't useful for cryptography.

RSA lost my trust when I found out that root keys or critical secrets for the SecureID system existed in the first place.

If they designed the system well, they would make physical tokens and deliver the tokens and the keys for those tokens to the clients. They would not keep any record whatsoever of those keys, nor would they have a way to reconstruct them

As a physical analogy, a locksmith making safety deposit box keys should make a random key and not write down the key code

Before the battery replacement, the car lost 3.5kWh/day. After the replacement, it lost 1.1kWh/day. That's a difference of 2.4kWh/day, which is 100W. That's something like 8 Amps internally leaking in the 12V battery. That seems shockingly high. Or maybe there's something else going on. If the battery was marginal, then perhaps the car's DC-DC converter was continually "charging" it but actually overcharging it. Then it would be electrolyzing 8A worth of water and battery acid. I expect that would make a giant mess. Alternatively, it just keeps running the DC-DC converter at very low output. The DC-DC converter could be incredibly inefficient when producing just a little bit of current (Tesla is reputed to use a huge (~200A) DC-DC converter, so the thing could be running at about 1% of rated output at, say, 10% efficiency).

Or we could finally fix the law and declare EULAs to be unenforceable. Unilateral contracts like EULAs are out of control.

A class action lawsuit for gross negligence might do the trick.

Sometimes I think that things like this should be felonies, though. Criminal offense or not, in a sensible world this would put alphanetworks out of business.

That's not the only problem. Doing this switches you from using a well-standardized, well-supported language to a nonstandard subset. This removes a lot of the advantages of Java.

One of the most technically competent exchanges I work with is trying out Azul, which is (AFAIK) the top-of-the-line low-latency JVM. I was a bit surprised to learn that their software didn't Just Work. In fact, it didn't work at all. So now they have to decide which AOT-compiling JVM to run and figure out how to support it. I suppose that Excelsior and gcj (eek!) would be other options.

My main point here is that, when low latency is involved, a lot of people cite the existence of high-quality real-time JVMs as evidence that Java is a good choice. These JVMs exist, but they are by no means panaceas, and users of Java for real-time or low-latency work will need to make a real technical and monetary investment in dealing with them. Hotspot is not (yet?) appropriate for this kind of thing. Oddly, .NET seems be in better shape here, at least in terms of AOT compilation.

I'm not a Java user, so I've never directly tuned for things like GC, nor do I interact with it directly. Warm up is a different story.

I interact with quite a few exchanges (over all kinds of protocols). Most are, unsurprisingly, written in Java. Almost all of them perform terribly at the beginning of the week. The issue is a standard one: the JVM hasn't JITted important code paths, and it won't until several thousand requests come in. For a standard throughput-oriented program, this doesn't matter -- the total time wasted running interpreted code is small. For a low-latency network service, it's a different story entirely: all of this wasted time happens exactly when it matters.

The standard fix seems to be to write apps to exercise their own critical paths at startup. This is *hard* when dealing with front-end code on the edge of the system you control. Even when it's easy, it's still something you have to do in Java that is entirely irrelevant in compiled languages.

If JVMs some day allow you to export a profile of what code is hot and re-import it at startup, this problem may be solved. Until then, low-latency programmers should weigh the faster development time of Java with the time spent trying to solve annoying problems like warm-up.

No, because the summary is (as usual) thoroughly overstated. This experiment, like any other form of quantum state tomography lets you take a lot of identical quantum systems and characterize them. For it to work, you need a source of identical quantum states.

As a really simple example, take a polarized light source and a polarizer (e.g. a good pair of sunglasses). Rotate the polarizer and you can easily figure out which way the light is polarized. This is neither surprising nor a big deal -- there are lots of identically polarized photons, so the usual uncertainty constraints don't apply.

The whole point of QKD (the BB84 and similar protocols) is that you send exactly one photon with the relevant state. One copy = no tomography.