Okay. BGP hijacking would be valid. Compromising the server would mean that you do have effective control over the domain and you don't even need to do a man-in-the-middle as you'd have access to the un-encrypted data anyway. I'm not sure what you mean by poorly secured DNS services - the challenge/response is performed by LetsEncrypt so DNS poisoning the client wouldn't work very well. You'd have to DNS poison the LetsEncrypt servers which would pretty much be a BGP hijack.

So how does the client machine know about your self-signed CA? Couldn't a man-in-the-middle attack do exactly the same thing and your client wouldn't know (if it was the first visit) whether it was your self-signed CA or someone else's self-signed CA?

I was genuinely curious, so thanks for your answer.

If I'm reading your post correctly, it sounds like a BGP attack. I just did a quick check and found that LetsEncrypt does also check previously issued certificates, so if you've already got a (valid) cert for your domain, then it's going to be a lot harder to get LetsEncrypt to issue a cert to the spoofed domain (the attacker wouldn't have the original private key).

The biggest issue with self-signed certificates is that the client machine cannot verify if the certificate belongs to the domain owner. If you're running a malicious wifi spot, you can do a bit of DNS poisoning to direct your clients to the wrong IP address and then present a different self-signed certificate and perform a man in the middle attack.

A LetsEncrypt cert can only be issued to someone who controls the domain in question and so gets round the man-in-the-middle issue.

How would a man-in-the-middle get a LetsEncrypt cert for a domain that they have no control over?

I've seen MITM attacks at several wifi hotspots. Airports are a particular favourite place for people to set up a rogue hotspot and grab loads of credentials. To be honest, the safest way to use hotspots is to encrypt everything by using a VPN, but at least HTTPS will give you some warning (invalid certificates) if you do connect to a rogue hotspot without using a VPN.

I'm not understanding the "vetting" issue with LetsEncrypt - they don't do anything except determine that you have control of the domain. It's automated, so there's no-one deciding whether or not to issue a cert. Even if they did refuse to issue a cert, then you can get one from elsewhere.

I don't really get why people are against encryption - can you clarify "LOT of downsides" for me?

Nope. A simple analogy would be to use the postal system. Imagine that HTTP is like people sending each other postcards. Anyone can read them whilst in transit and also alter them. HTTPS would be equivalent to everyone sending letters in sealed envelopes (maybe with old-time wax seals on them). Now I understand that you don't want to be funding the BIG envelope corps, but here's an initiative that provides free envelopes (although they bio-degrade after 90 days which some people think is awkward).

The biggest problem with HTTPS is that it uses more CPU and prevents caching, but CPU usage isn't really a problem these days.

Self-signing is easy enough but has security issues. The client has no way to determine who did the signing - it could be the website owner or it could be a man-in-the-middle.

Using HTTPS everywhere is more about protecting client computers (and their data) rather than needing a third-party's blessing. LetsEncrypt is a major step in lowering the barrier to let everyone run HTTPS easily and for free. It's designed to be easy to automate, so all you have to do is set up your web server to allow the specific challenge/response mechanism to verify that you have control of the domain. Then a one-line command is all you need to get your certificate in seconds. Point your webserver at the new certs and away you go. Renewal uses the same challenge/response system, so you just leave the relevant section in your config and you're ready for automated renewal.

In general, that kind of page doesn't need to be encrypted.

However, encrypting connections to websites makes it harder for bad guys to sabotage someone's connection to the website and injecting malware/ads etc. A free and easy to get and use SSL cert provides some protection for very little cost, hence the push to get as much of the web encrypted as possible.

There's also an issue where people might be trying to analyse traffic and it could be of some advantage for them to know when you're visiting "secret encrypted" sites versus "ordinary http" sites. Encrypting everything by default can hope to allow some anonymity.

I prefer the break-early model of LetsEncrypt. Set up your test system with free LetsEncrypt certs and then test the cron script (one-liner) for renewing. Also, the certbot client has a dry-run feature so you can check what it's going to do if you do want to do proper testing.

With long expiry dates, you'll never get around to automating renewal and then you'll probably forget all about it and/or move to a different job and not care. Someone is then left with a ticking time-bomb of embarrassment for a domain cert running out and probably no available test system (oh, that service hasn't been touched since Fred left - no we don't know how to re-create it for test).

For me, NoScript provides decent protection against that. If I've visited the correct site previously, then I would've white-listed JavaScript for that domain (and possibly a couple of their related domains). Then, if I visit the fraudulent domain, no JavaScript would run and the chances are that the site would look very different.

SSL certs are not primarily for identifying that you've visited the correct domain (as in the one that you think you were connecting to), but are to prevent man-in-the-middle attacks and ensure end-to-end encryption.

You mention "(unproven) theory", whereas the scientific use of the word "theory" signifies that it is as close to proven as possible i.e. no data has so far contradicted it.

Another point regarding these microscopic black holes is that cosmic rays behave like extremely powerful particle accelerators in that particles are smashed together with far greater energy than we can manage in our own accelerators. If microscopic black holes are a problem then we wouldn't be here to complain about them.

His parents were Libyan's what?

If you could copyright your stupidity, you'd be rich.

Here's some of the reasons why your analogy is invalid.
Ford produces cars - Elsevier does not produce the papers.
Ford cars are not funded by the tax-payer whereas the Elsevier papers are.
You wouldn't download a car, would you? (Cars are not copyrightable - they are a physical good)

I've had a quick look at UK (where I live) statistics (e.g. https://www.gov.uk/government/...), but it's not much help as we've had reducing number of pedestrian fatalities since 2004 (excepting 2011 which is blamed on heavy snow).

As we have plenty of smartphones and plenty of smombies, that indicates that there's something else happening in the U.S. that's increasing fatalities, but it'd be premature to blame smartphones.