awelkie

Could you be more specific? What about this proposed "ideal" scheme is infeasible (specifically the engineering limitations you mentioned)?

If your ISP didn't have that feature, could you just replace the cable modem too? My ISP's router is running EuroDOCSIS 3.0 and I'm wondering if I could replace the router with a modem + router of my own.

This seems like it could be useful for UWB communications or a very precise radar. Is it easy to generate an EM impulse with similar bandwidth? I guess the difficulty is to do all signal processing in the optical domain, since it's impractical to digitize all of that spectrum.

Excellent point. So how does Google solve the problem? Do they just coordinate the heartbeat intervals with the carriers, or do they bypass the Internet altogether and use some cell-network-specific push mechanism? If the latter, is this mechanism available to everyone or just partners of the carriers?

Great article; very interesting and well written.

A bit of a tangent, but something that bothers me is the assumption in economics that the amount of money one is willing to pay for a good is equal to the utility provided by that good. And then it follows that if I am willing to pay €2 for a coffee and another person is only willing to pay €1 for that same coffee, then I value the coffee twice as much as the other person.

But to me that doesn't make much sense. What if I am a billionaire and the other person is very poor? It could very well be that the other person values the coffee a lot more than I do, but because I have so much more disposable income than they do I am willing to pay more for it. To me, the assumption should be something like the amount of money as a percentage of my total wealth that I'm willing to pay is equal to the utility of the good.

And because this assumption is so fundamental to economics, it determines the conclusions that the field reaches. For example, the definition of an "efficient market" or the optimal level of production of some good both embed this assumption that willingness to pay equals utility. And then one can imagine that policy decisions and such are influenced by this assumption. (And in fact the article itself mentions this problem: "Another challenge is that quadratic payments, being a payment-based mechanism, continues to favor people with more money.").

So I'm wondering if anyone has looked into an alternative formulation of economics with a different fundamental assumption (perhaps something closer to the one I proposed above). If so, do certain things that are determined to be "optimal" in classical economics turn out not to be optimal in this alternative formulation (and vice-versa)? Apologies if this is a bit of a naive question; Econ 101 is the most I've ever studied the subject.

In fact, the French have separate words for the stimulants in coffee and tea: caféine and théine. And if your goal is to describe the "source of stimulation" then this may be more accurate than using the same word for both coffee and tea as the stimulating effects come from a variety of components in coffee and tea (as a sibling comment points out).

Also, hi Stefan!

LoRa is just a PHY, LoRaWAN is the MAC. And last I checked LoRaWAN used ALOHA, not slotted-ALOHA, which is even worse. But I knew a few years ago they were talking about adding beacons to the MAC to enable slotted-ALOHA, so maybe my knowledge is out-of-date?

The problem is two-fold. First, LoRa is a spread spectrum technology that also has a problem with collisions. Spread spectrum is where the signal bandwidth of your transmission is higher than the information bandwidth, meaning your transmission's spectral efficiency is terrible. But usually spread spectrum systems are designed such that multiple transmissions can occur at the same time without colliding, so the network spectral efficiency is just as good as a narrowband system. But LoRa is a spread spectrum technology that still suffers from collisions, so the network spectral efficiency is really bad.

Second, LoRaWAN uses an ALOHA medium access scheme, which also suffers from poor network spectral efficiency compared to basically every other medium access scheme out there. So these two things combined mean that a LoRaWAN network probably won't be able to handle much traffic.

Almost all of the media I see about LoRa and LoRaWAN talk about the great range, which is really good, but then gloss over the network performance. So in my opinion LoRa is a bit too hyped at the moment. I also think that when LoRa came out, it was the only PHY that bothered to go down to really low bit-rates, and then the range basically comes from the Shannonn-Hartley theorem. People think that LoRa gets its range from some fancy signal processing tricks or whatever, but I think it's just the fact that they bothered to drop the bit-rate very low. That's basically what SigFox is doing as well, but they're using a more typical narrowband PHY and so probably won't suffer from the same network capacity problems.

I think that's what the poster is saying. I believe "Cannabis causes schizophrenia!" should be in quotes.

You're right of course. With Mel I know he's very forthright about his career in the NSA, so I figured it would be alright. I just confirmed with him too, just to make sure.

If this was before 2015, there's a chance the person giving the talk about inclusivity was my father-in-law; I know in the last few years of his career there he got involved in the push for diversity. He always talked very fondly of his career at the NSA.

I think in a lot of ways the NSA is a better workplace than any silicon valley tech company; you don't really have to worry about profitability, there's an enormous breadth of interesting work to be done, and you get to work with a lot of really talented people (I think the NSA is the largest employer of mathematicians in the US). Of course there are downsides too, like the low pay (set by Congress) and the constant drug tests and polygraphs.

Based on my discussions with him, I believe that the organization has two conflicting goals; to improve the IT security of the US and its allies, and to weaken the IT security of everyone else. And there are historical examples of the NSA doing both. But internally apparently there is a lot of debate about what the NSA should be doing, especially post Snowden. So yeah, I can believe that plenty of people at the NSA are deservedly proud of their work. Not everyone there is a cynical government drone working to undermine IT security globally. But of course when the NSA starts contributing to your project, you don't know which of their two goals they're working towards...

As an aside, my father-in-law is a very passionate mathematician, and in his retirement he just published a book on some interesting and approachable topics in mathematics that much of the HN crowd would probably find interesting: https://bookstore.ams.org/mcl-22/

Furthermore, were cryptocurrencies to become illegal, it would probably be similar to computer viruses; It's fine to play with viruses on your personal computer or in a lab, but it's illegal to use one "in the wild." So I doubt it would be as absurd as you claim.

>If this legislation passes, any computer scientist, mathematician, programmer who is working with [automated ponzi schemes] can technically be held liable for [creating a ponzi scheme] when they run their program?

I don't see the issue. When a democratic society thinks something isn't good, they can outlaw it. There are all sorts of regulations on financial transactions and investments already.

I'm not trying to imply that cryptocurrencies are ponzi schemes, just that when something is illegal, doing it on your computer doesn't make it not illegal.

It's very annoying. Especially since the reason I use gnome-characters is because I'm in the middle of typing something and I need a character that's not easy to type on my keyboard, so it's already a distraction from what I'm trying to do.

As an aside, I've been trying out Ubuntu for the last 6 months, and this is one example of the lack of polish I'm seeing. I don't remember installing gnome-characters explicitly (it's possible I just forgot doing so), but there are two versions on my machine, one of which happens to have this crazy startup lag. Another annoyance is that firefox came installed as a snap, which meant that all of my downloads went to something like ~/snap/firefox/Downloads instead of the expected ~/Downloads. I don't remember these kind of things happening when I ran Fedora...

Is that the first time per installation, or the first time per boot?

I just tried starting the gnome-characters app, which I have installed via snap and via apt. The snap-installed app boots in around 6 seconds after restarting my machine, and then in about 3 seconds afterwards (I'm just counting the seconds in my head). By contrast, the apt-installed app boots well under a second.

You could do something smarter with error correcting codes. Using a rateless code you split your message into an infinite number of chunks, then send these chunks out on your multiple paths, and then once enough chunks have been received you ACK each path. No bandwidth wasted.

I think specifically with voice it should be possible to send two chunks such that if they both arrive then you get your audio, and if only one arrives then you still get your audio but at a lower quality.

Not to get too off topic, but I have a hard time blaming the left for the current political divide in the United States. Saying that someone who pushes leftwards is bad for American politics because they're not closer to the "center" seems like promoting a false equivalence.

I think it depends on the field. In mathematics I think the expectation is that students need to study for a few years to even understand what the open questions are, and only then can they begin the research.

I'm fully aware of what the point of a PhD is. I wasn't expecting to become a generalist, but I was expecting some time to study some recent results and come up with a tractable research proposal. Indeed, if the purpose of a PhD is to expand the totality of human knowledge, why was it expected that I would publish 3-4 conference papers to graduate? Wouldn't one significant result have been enough?

I think CS PhD programs in particular can be a little strange. I decided to go back to school to pursue a CS PhD after chatting with a retired mathematician about his PhD. He was talking about how he took classes for two years to explore different areas of mathematics and to pass the general exams, then he decided on a branch of mathematics (topology) and approached a professor in that area to be his advisor. After working through a few textbooks together, this person posed a question to his advisor, spent a few more years working to answer the question, eventually answered it and published the result in his dissertation.

This seemed to be what a PhD was all about, but it's not what I've experienced. I've just completed the first year of grad school, and while the classes have been interesting, the expectation is that you already have an advisor when you arrive (or at the latest at the end of your first year) and you hit the ground running pumping out papers. There seems to be little opportunity to explore computer science as a whole or to work towards one singular result. I recently had the pleasure of chatting with Peter Sarnak about CS graduate education. He said that CS is in a weird spot between traditional engineering disciplines and the liberal arts, and that the need for grant funding causes departments to encourage students to work on whatever projects the professors are leading. But even he seemed a little surprised when I told him that our department's general exam is basically a presentation on the research you've accomplished so far.

I think this is unfortunate. First, because it goes against what I wanted out of a PhD. But also because it makes the PhD not much different than working in industry. If in either case I have to write code for some boss and satisfy my curiosities on the nights and weekends, why would I choose to do it for much less money and worse career prospects? I imagine this is causing the best and brightest to avoid academia in computer science, which seems bad for research and technological progress in general.

FreeRTOS is essentially a unikernel. I assume many other RTOSs are as well.

Oh yeah, that's an interesting solution.

I'm not sure simultaneous authorization and anonymization is impossible. Couldn't you use something like Chaum's e-cash to obtain tokens that guarantee the holder the right to use the network for some amount of data, but these tokens are tradeable and therefore the spender doesn't have to be the same as the buyer. Then you could spend this token in the network to get access and the network could authenticate the token without identifying the spender. I'm guessing something like zcash could be used as well...

Calls could be done over IP, and as long as you could anonymously authenticate to the tower then you could be granted a new IP address at each tower via something like DHCP. I imagine roaming and handovers would have to be done on the end-device though; the end-device would need to proactively associate to new towers and both ends of the voice call would need to agree to switch to the new IP address.

But if the tower operators collude then they can still track you across towers by localizing the physical source of the end-device's signal.

I don't think it's possible through technological means to avoid being tracked and still use a wireless network. Even if you could anonymously authenticate to the network, if the base stations have a large number of antennas then they can locate the physical origin of your signal and track you that way.

It may be possible of course through other means, like government regulation or only using carriers that have some guarantee of privacy.

You may be thinking of this article: https://news.ycombinator.com/item?id=16881946

Good point about owning vs leasing. But there's also a lot of room between private spectrum rights and the ISM band free-for-all.

Also, in the paper they say they used pulse-width modulation for the backscatter signal instead of just attaching the sensors directly to the antenna like the Great Seal Bug because the dynamic range of the photodiodes is much less than that of a microphone and they wanted to avoid using an amplifier (section 3.1 in the paper). But the PWM requires power as well. Would an amplifier that only has to map the range of a photodiode to the range of a microphone really require that much more power? And if so, are we talking like twice the power, 10x, 100x?

Very cool project.

I wonder if compressed sensing would be useful here. In the paper they describe this custom (and seemingly ad-hoc) video compression algorithm that is suitable for the low-power device. Basically, the camera averages blocks of pixels into "super pixels" all in the analog domain, sends these super pixels over the wireless link, and then the receiver can request the individual pixels of any super pixel that seems like it might have changed since the last frame. But with compressed sensing, can't all of this be done by sending random linear combinations of all pixel values (like the single pixel camera)? It seems like then you can avoid the bi-directional link and also get rate adaptation for free.

Definitely, this project is basically the Great Seal Bug but for photodiodes. The authors even mention the Great Seal bug in their paper.

There has been some recent work[1] towards backscatter video (actually from the same lab you linked to). The basic idea is the same as the "great seal bug"[2] except for photodiodes. I say "towards" because much of the work they present is in the form of simulation.

[1] https://homes.cs.washington.edu/~gshyam/Papers/videobackscat... [2] https://en.wikipedia.org/wiki/The_Thing_(listening_device)

I feel like there's an opportunity for a web platform for the collaboration of (independent) researchers. Some place where I could propose an idea and others could tell me why that idea doesn't work or if it's already been done. And if the idea doesn't have any flaws and hasn't already been explored, then other researchers could help me perform the research or "fork" the project to go down different paths.

I think some platforms come close. There are places to ask questions (e.g. StackExchange) but they don't seem to like "what if" questions and if the idea is good there's no good way to follow up over months or years. Github works for some fields where the idea is tightly coupled with the implementation (e.g. some computer systems fields) but not for others.

I guess I would just like a place to discuss ideas. In academia it's common to discuss nascent ideas within your lab, but larger collaboration happens at things like conferences where you're presenting only the ideas that worked out. I think such a platform could make independent research like you're describing much more effective and attractive.