ash

Thank you for summarizing my assumptions, but it was not fully correct:

- I'm not saying backdoors are "routinely" detectable, I'm simply saying it's possible to detect them. In contrast, we have zero possibility to find out if Telegram is doing something nasty with your chats. Telegram may well do it already, and we may never learn.

- Regarding the incentives - security researchers often disclose vulnerabilities to the public. Obviously they have the incentives to do it.

As for the optional nature of Telegram E2E encryption - Signal and WhatsApp successfully demonstrated it's possible to make E2E the default (and only) option. There's no competition here. I wonder what proportion of Telegram users understand the ramifications. Many people migrating from WhatsApp to Telegram probably don't.

We don't need to play any games to point out that Telegram can see everything, including messages.

E2E encryption is necessary for all communication. Email is unfortunately not encrypted, but it doesn't explain why would you choose non-E22 tech when there are perfectly good E2E alternatives.

E2E is an insurance against:

* bugs

* rogue employee

* server vulnerabilities

* acquisitions (imagine Telegram being acquired by Facebook)

As for metadata analysis, I still fail to understand how is it comparable to Telegram having access to all messages.

Sure, WhatsApp can go rogue at any time. Signal app is obviously better. However, at least we have a possibility to find out if WhatsApp goes evil - decompilation, protocol analysis are still available. I wouldn't also discount fear of reputation losses if E2E vanishes from WhatsApp.

Compared to maybe-evil-in-the-future WhatsApp, default Telegram is already evil, because they have access to all communication right now. What do (or will) they do with all that data?

(Opt-in E2E - and only for 1-on-1 chats - is almost useless. Encryption should be the default.)

Closed source WhatsApp is a problem. And a backdoor is possible. However, it's still detectable, albeit not easily. I'm sure researchers are closely looking at WhatsApp binaries. WhatsApp backdoor would be a scandal.

Sure, open source WhatsApp would be better. However, we are comparing detectable potential backdoor with totally undetectable existing access to all non E2E chats on Telegram servers. Telegram developers can already see everything right now. (By the way, opt-in E2E encryption is almost useless. Encryption should be the default, and enabled for group chats too.)

Between these two options WhatsApp situation is clearly better.

The author of the parent comment specifically mentioned "change in privacy update" as the reason to switch. Until WhatsApp drops end-to-end encryption, there's not that much evil they can do. They can't read your messages. You don't need to trust WhatsApp (so far).

Telegram can have the best privacy policy, but they can always read your messages. You have to trust Telegram to do the right thing. (Except those few who opt in to use encrypted chats.)

Strange to see people switch from end-to-end encrypted WhatsApp to non-encrypted (by default) Telegram. What do I miss?

This was helpful:

https://en.wikipedia.org/wiki/IPv6_address#Scoped_literal_IP...

Apparently it's possible to embed interface name into an address:

fe80::1ff:fe23:4567:890a%eth2

The article also points to an alternative solution - embed interface index into an address:

fe80:3::1ff:fe23:4567:890a

One answer to "who?" question: https://news.ycombinator.com/item?id=26419035

Still unclear why and where do zones come from.

> Especially considering how rare IPv6+zones are.

Who uses IPv6+zones and why?

I'm confused what zones are about. They don't seem to be part of IPv6 packets. Where do they come from? Why does IP type have to deal with them?

> The function color problem is a bit overblown, it hasn't broken nodejs yet either.

It didn't break Node because Node avoids function color problem altogether. Nobody writes blocking code in Node due to its single-threaded nature.

Function color problem is the difficulty of mixing blocking and async functions. It's especially annoying in Python, where you have to spawn a thread to call blocking functions from async functions, and welcome to mutithreading bugs.

In Node the problem doesn't exist, because you don't block in its blocking functions. You either do `await f()` or `f().then()`.

Further down the thread: https://news.ycombinator.com/item?id=26407770

Is there a good explanation on the difference between polling model and completion model? (not Rust-specific)

I think I got it now!

I was confused because Tailscale does not bring its own userland TCP/IP. It can - as a VPN solution - rely on OS-provided TCP/IP stack, but you wanted to avoid having to hook up flyctl into OS as a virtual network interface, right?

In general, I dislike the extra complexity JSON5 brings.

For example, allowing identifiers as object keys.

Why is userland TCP/IP stack needed? I didn't get this part of the story.

I find that JSON5 does too much.

Sorry about the title! It was my (probably poor) attempt at putting an exciting spin on the news. In the sense of "all publicity is good publicity".

Dark is a very cool project. Every time I wait for a long CircleCI job to complete, I wish our programming environment was like Dark.

Good luck with it!

One of the commenters gave a great suggestion:

> I would use bar models to show how the whole changes. First, draw one bar (table) with 3 students inside; label the bar one whole, star one student and label that student as one third of that whole. Do the same thing beside the first bar model, again showing the bar as one whole, starring one student and labelling that student as one third. Then, push the two bar models together (I’d use a Smart Board) to show a new whole: the two bars together with 6 students in the one bar is labeled as the new whole. Point out that this is a NEW whole, with a different # of pieces. Now they would see the 2 starred students in the one new bar made up of 6 total students, so the “old” 1/3 student becomes the “new” 1/6 student once the whole changes. This is like a name change, once the size of the whole changes. You can also show them that by cutting each “student” in half, you would get an equivalent fraction of 1/3 = 2/6 of the first one whole bar. So 1/3 = 2/6, not two 1/3’s = 2/6.

http://www.marilynburnsmathblog.com/can-1-3-1-3-2-6-it-seeme...

> Li-metal batteries top at around 14 MJ / kg

Do you have a link to back this number? Are these batteries rechargeable? Does the measurement include wiring, cooling, cell connections and overall structure of the battery pack?

For comparison, Tesla's (Li-ion) battery pack energy density is 160 Wh/kg, which is 0.576 MJ/kg:

https://www.pv-magazine.com/2020/02/11/energy-density-advanc...

> [1, 2, 4, 8] is exponential and [1, 3, 5, 7] is linear.

You can't say just from 4 data points. Real data is noisy. Imagine each number is ±1.

Awesome feature!

Is there a way to change resolution during editing? When I zoom in in Photopea, I can see individual pixels of text characters. With some PDFs pixels show up sooner, making text barely readable during editing.

Very recent (March 28, 2020) non peer reviewed preprint:

"Correlation between universal BCG vaccination policy and reduced morbidity and mortality for COVID-19: an epidemiological study"

https://www.medrxiv.org/content/10.1101/2020.03.24.20042937v...

I wonder if there's research comparing Covid-19 deaths versus BCG Tuberculosis vaccination:

1. Between countries. E.g. Russia (mandatory vaccination) versus Western Europe (canceled some years ago).

2. Between territories. West Germany versus East Germany.

3. Between people. Migrants to US from mandatory vaccination countries versus US locals.

More details from the tweet author: https://forbetterscience.com/2020/03/26/chloroquine-genius-d...

Without paywall: http://archive.vn/cgCff

Total number of confirmed infections is a misleading metric. Criteria for testing constantly changes everywhere. Test kits become more (or less) available. Hospital resources are allocated differently based on changing situation, etc. Two random examples:

* March 15th, Southern Finland - "Coronavirus testing to focus on health professionals. … people returning from trips abroad will not necessarily be tested, let alone all of the people suffering flu symptoms." https://yle.fi/uutiset/osasto/news/hus_coronavirus_testing_t...

* UK NHS now recommends simply staying at home and not even contacting them in mild cases. https://www.nhs.uk/conditions/coronavirus-covid-19/self-isol...

Italy, China, U.S. most likely determine cases the same way so far. No wonder CFR is similar. However, these tests are limited compared to serologic tests:

> The test kits in use in the U.S. described in the articles detect viral genetic material—RNA, in the case of coronaviruses—which can be infectious material or noninfectious fragments. Once the patient has recovered and the RNA has been cleared, the tests will be negative. If we’re trying to ascertain what proportion of the population has been infected and experienced asymptomatic, mild or more serious infections, such post-infection testing yields “false negatives.”

> Additional, essential information will need to come from “serological tests” that measure antibodies in blood, which will tell us whether a person has been recently infected with SARS-CoV-2 and recovered. (Note that antibodies take about 10-14 days from exposure to the virus to appear.)

https://www.wsj.com/articles/do-antibody-testing-to-understa...

http://archive.vn/SJkJu

> The mortality rate is thought to be ≤2%, but precise numbers uncertain due to a lack of available serological testing.

https://www.hopkinsguides.com/hopkins/view/Johns_Hopkins_ABX...

Another source:

> "The current PCR test only tells you whether a person has the virus at this moment," Dr Meru Sheel, an epidemiologist and research fellow at the Australian National University said.

> "What it doesn't tell you is that a person may have had the infection, has recovered and is immune or not immune."

> Dr Sheel said a serological test would help understand at population level how many people have been exposed to the disease, have recovered and if there are populations that are not immune to the virus.

https://www.abc.net.au/news/2020-03-17/coronavirus-experts-c...