Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Improving first impressions on Signal

signal.org

Post reply on HN
Print view View original

41–50 of 100 posts

Re: Improving first impressions on Signal

#41

Post by Tistron »

Is there a reason, when they have control of the client software and the protocol, that they don't introduce computational cost to establishing contact? Like, if I send a message to a particular number, and their client receives a message from me for the first time, it responds with "prove that it's worth it for you to message me: find an `n` character string that together with `random string` hashes to 5 leading zeroes", or something like that. Somehow I imagine that doing a few seconds of computation per initiated first message is unproblematic, but doing it for thousands or millions of numbers starts to be a problem.

Why is this not done?

Re: Improving first impressions on Signal

#42

Post by pocketlim »

modeless wrote: post #21

Just got a new phone and migrating Signal was a total disaster. The first thing I did was login to the new phone which is apparently the wrong thing to do because you lose all your old messages and there's no option to import after you login. You have to logout but there is no logout . Gotta uninstall and reinstall, or clear app data. Next I tried the account migration feature which requires using both phones simulta…

Recently got a new phone and the transition was super smooth. Tried to open Signal on the new phone, it somehow knew that I needed to transfer my backup, asked me to scan a QR code and keep the phones close together, a minute or so later, and done. Signal logged me out on the older device when it was finished.

I've learned this lesson before though. Always keep your old phone around while transitioning to a new device in case you need to pull anything off of it.

Re: Improving first impressions on Signal

#43

Post by tptacek »

Tistron wrote: post #41

Is there a reason, when they have control of the client software and the protocol, that they don't introduce computational cost to establishing contact? Like, if I send a message to a particular number, and their client receives a message from me for the first time, it responds with "prove that it's worth it for you to message me: find an `n` character string that together with `random string` hashes to 5 leading zer…

If the server is mediating that proof-of-work system, then the server is building a list of contacts, and not doing that is the #2 goal for the whole Signal project.

If the client is doing it, you need to come up with a system that is resilient to people using a bunch of different devices, getting new phones, etc.

This is a theme about building Signal: they're doing everything on hard mode, because they start from the premise that they don't get to know everyone's contacts. Virtually every other mainstream messaging system, including the ground-up E2E encrypted ones, keep a complete plaintext contact database serverside.

Re: Improving first impressions on Signal

#44

Post by Tistron »

tptacek wrote: post #43
Tistron wrote: post #41

Is there a reason, when they have control of the client software and the protocol, that they don't introduce computational cost to establishing contact? Like, if I send a message to a particular number, and their client receives a message from me for the first time, it responds with "prove that it's worth it for you to message me: find an `n` character string that together with `random string` hashes to 5 leading zer…

If the server is mediating that proof-of-work system, then the server is building a list of contacts, and not doing that is the #2 goal for the whole Signal project. If the client is doing it, you need to come up with a system that is resilient to people using a bunch of different devices, getting new phones, etc. This is a theme about building Signal: they're doing everything on hard mode, because they start from th…

Thanks.

What is the problem with "people using a bunch of different devices, getting new phones, etc" in what I suggested?

If I make a fresh install, then next time a friend messages me, it costs them a few seconds of delay in sending their first message because my client asks them to prove themself.

I don't see the problem, but I imagine there are problems since smarter people than me are not doing it this way. I'd like to understand what I am missing here.

I guess tweaking the difficulty to be actually difficult for a server farm, yet possible for an old device would be the main problem.

Re: Improving first impressions on Signal

#45

Post by tptacek »

Tistron wrote: post #44
tptacek wrote: post #43

Earlier quoted context omitted.

If the server is mediating that proof-of-work system, then the server is building a list of contacts, and not doing that is the #2 goal for the whole Signal project. If the client is doing it, you need to come up with a system that is resilient to people using a bunch of different devices, getting new phones, etc. This is a theme about building Signal: they're doing everything on hard mode, because they start from th…

Thanks. What is the problem with "people using a bunch of different devices, getting new phones, etc" in what I suggested? If I make a fresh install, then next time a friend messages me, it costs them a few seconds of delay in sending their first message because my client asks them to prove themself. I don't see the problem, but I imagine there are problems since smarter people than me are not doing it this way. I'd…

I'm not saying it isn't potentially workable, just that it's deceptively hard to deploy across the whole Signal userbase. A company like Slack or Discourse could build something like this easily, because it'd be mostly managed from the servers, which they control, working from a single source of truth about who's talking to who. Signal doesn't have that luxury.

Re: Improving first impressions on Signal

#46

Post by teawrecks »

ISL wrote: post #3

Is it strictly-true that anti-spam algorithms must be hidden to be effective? It would seem that initially, spammers would have the upper hand, but in the long run, coordinated and open effort against spammers has the potential to pay off. Furthermore, to have a closed algorithm that cuts off the ability for people to communicate opens the door to institutionalized censorship -- designating a political opponent's com…

It's the same as anti-cheat in a game. Given how easy it is to create a cheat for open source code, it's just way more cost effective to keep it closed.

The rule is not to RELY on security by obscurity. It shouldn't be the last line of defense. But by all means, use it where the ROI makes sense.

Re: Improving first impressions on Signal

#47

Post by aeturnum »

modeless wrote: post #21

Just got a new phone and migrating Signal was a total disaster. The first thing I did was login to the new phone which is apparently the wrong thing to do because you lose all your old messages and there's no option to import after you login. You have to logout but there is no logout . Gotta uninstall and reinstall, or clear app data. Next I tried the account migration feature which requires using both phones simulta…

I do think it's too easy to do it the Wrong Way, but I've easily migrated from phone to phone several times now with both techniques and think they alright.

Signal presents both a "create account" or a "migrate account" option on a fresh install, which seems like the best compromise for UX flow.

Also, considering their guarantee is to keep your messages secure, of course you need your old phone. If you wipe your phone that securely contains all your old messages they are gone. That is security.

That said, if you have the backup of your messages and the code, then the developers might be able to debug what caused the migrations to fail if you submit a bug.

Re: Improving first impressions on Signal

#48

Post by kemayo »

noja wrote: post #14

Great, but... Please let people allow Signal to automatically save their media to their photo roll (or whatever ios calls it). The number of people I have met who require this and left Signal _after starting to use it_ is astonishing.

I mean, I'm not opposed to this feature existing, but I always thought it was one of WhatsApp's weirder choices as a default behavior.

I certainly turn it off the moment I set up WhatsApp on any device -- why would I want random memes mixed into my photos?

Re: Improving first impressions on Signal

#49

Post by xboxnolifes »

ISL wrote: post #3

Is it strictly-true that anti-spam algorithms must be hidden to be effective? It would seem that initially, spammers would have the upper hand, but in the long run, coordinated and open effort against spammers has the potential to pay off. Furthermore, to have a closed algorithm that cuts off the ability for people to communicate opens the door to institutionalized censorship -- designating a political opponent's com…

> Is it strictly-true that anti-spam algorithms must be hidden to be effective?

No. But that "no" is missing some qualifiers. It's not strictly true since you can implement some out-in-open options such as paying money, requiring a phone number, validating your physical identity, delays of use on new account creation, etc, that will make the barrier of entry for spamming much higher. There will still be spam, but it'll become more manageable to manually monitor and take action against. The downsides I feel are obvious based on the service we are talking about.

But I would say it is strictly true that you cannot have a free, anonymous, low barrier to entry service without hiding your anti-spam algorithms. There is fundamentally no difference in such an environment to differentiate a bot from a human besides the message content and patterns in who they message. If bots know what content they can't send or what patterns in messaging to avoid, they can still spam.

Re: Improving first impressions on Signal

#50

Post by vlmutolo »

jakecopp wrote: post #36

> To keep Signal a free global communication service without spam, we must depart from our totally-open posture and develop one piece of the server in private: a system for detecting and disrupting spam campaigns. Interesting. Matrix solves this problem by not associating a username with a phone number or email address (unless you opt in to). Does anyone have views on if this will work into the future? I'm moved ~90…

How does not requiring a phone number help fight spam?
Post reply on HN