Comment by tptacek | Hacker News Reader

Comment by tptacek | Hacker News Reader

Parent Full thread

In fact, the Signal team has been explicit about why they won't federate, and, unsurprisingly, you're doing your level best to avoid being charitable to their explanation and to instead substitute your own reasoning.

Moxie's "interests" are in ensuring that that every aspect of the Signal ecosystem serves the goal of deploying cryptographic privacy to the largest number of people. This is why Signal didn't even have user profiles until recently, why it won't be until this year at the earliest that they have non-phone-number identifiers, and why even the GIF sharing feature of the application has its own purpose-built tunneling scheme. Signal simply takes privacy a lot more seriously than other tools --- especially the tool you yourself recommended, Matrix, which (while a fine and worthy project) didn't even enable E2E by default when you recommended it.

The problem federation poses to Signal's goal is the same problem Matrix faced: it forces the system into compatibility compromises with lowest-common-denominator clients. Right now, the Signal team can deploy a new privacy-preserving feature universally to all its users without negotiating with any third parties. With a federated design, they obviously could not do that. Since these kinds of features are essentially all Signal works on (much to the consternation of Hacker News), that issue is important.

But you know all this, because we've had the same "debate", with the same points made, ever since you wrote the "I Don't Trust Signal" article that derived from first principles Moxie Marlinspike's malign goals from the fact that Signal didn't support F-Droid.

The fact that you're not disclosing any of this context in your random snipes at Signal on unrelated threads is much better evidence of your bona fides than any of your arguments have been about Moxie's.

" This is why ... it won't be until this year at the earliest that they have non-phone-number identifiers ... Signal simply takes privacy a lot more seriously than other tools"

If they took privacy seriously they would have never asked for people's phone numbers.

No, the opposite is true, as you can see in other comments on this thread.

If by "other comments" you meant your own single comment which defended Signal's use of phone numbers by claiming "using phone numbers allows Signal to piggyback on people's local contact databases instead of maintaining that directory serverside", that does not prevent Signal from keeping a database of contacts serverside.

Further, they could have simply let users find and verify each other out-of-band, as PGP did with key distribution and fingerprint verification.

There is absolutely no need for Signal to know my phone number, and it requiring one has been the main reason why I haven't used it.

Using phone numbers doesn't prevent Signal from storing a database of who's talking to who serverside. Not using phone numbers would require them to do so, and, as you can see from other secure messengers that made different tradeoffs (like Wire), that's exactly what ends up happening.

It's pretty funny to try to use PGP as an example of a system that doesn't centralize and leak metadata.

How is PGP (now gpg) centralized? Out of band, I can exchange keys and verify fingerprints with whomever I want to communicate with, then send then a pgp/gpg encrypted email directly (or use some other means of sending the encrypted message). No centralization required.

As for Signal, there's no requirement for them to keep any kind of directory, with or without phone numbers. Just let people discover each other in some way not under Signal's control.

Whether other services have made the same mistake as Signal is irrelevant. It's still a mistake.

This point is so straightforward that at some point your inability to acknowledge it will become willful. It is simple:

People expect to find their acquaintances to talk to in a messaging application.

Most messaging applications solve this problem by keeping a serverside contact database. You see the same thing with Signal's competitors.

Signal refuses to keep that database, because it is practically as valuable to adversaries as message plaintext.

So, instead of having users register handles and maintaining a database, Signal uses phone numbers --- like the most popular messaging applications, including WhatsApp, iMessage, and SMS --- so that it can take advantage of contact databases users have locally and not need to store a contact directory globally.

"Just letting people discover each other some other way" doesn't work, and PGP is a good example of why: the "other way" PGP came up with is (wait for it) a global directory of key metadata.

You are free to disagree with the tradeoff Signal made here. What you cannot reasonably do is suggest that there was no tradeoff, or that their use of phone numbers is somehow obviously flawed and privacy-hostile. It is, as I said, the opposite: phone numbers are an instance where Signal sacrificed uptake and UX in favor of privacy, because they do not care what random people on message boards have to say about their decisions as much as they care about the fact that they're the world's most important and effective secure messaging design.

"People expect to find their acquaintances to talk to in a messaging application."

That hasn't been the case in my use of Skype, for instance. I'm not sure if it's changed, but when I used it it had no real name requirement, and the people I communicated with over it had to tell me their Skype usenames out-of-band, like through email or by phone. There was no directory that would have been useful to us, as we didn't use our real names (or phone numbers).

So your assertion that people on messaging applications expect to find their acquaintances (through it) is simply false in the case of Skype, which is one of the most popular messaging applications in existence.

Discord is another extremely popular messaging application that didn't require me to give my name or phone number.

"Signal refuses to keep that database, because it is practically as valuable to adversaries as message plaintext."

And a database of phone numbers is not?

"the "other way" PGP came up with is (wait for it) a global directory of key metadata."

If you're talking about pgp/gpg keyservers, those aren't necessary for communication via pgp/gpg, as keys can be exchanged and verified by other means. Furthermore, the entities (plural) most people communicate through using pgp/gpg (ie. mail servers) don't control the keyservers, so they're not centralized like Signal is. This is a distributed system.

"What you cannot reasonably do is suggest that there was no tradeoff, or that their use of phone numbers is somehow obviously flawed and privacy-hostile."

I can't believe that you can't see that requiring people to hand over their phone numbers is a violation of privacy, or that it could in fact be objectionable to people who care about their privacy. It's even more mind boggling that you'd argue that handing over one's phone number offers more privacy than not doing so.

This is akin to Facebook's real name policy, and arguably even worse, as so much tracking is tied to one's phone.

> That hasn't been the case in my use of Skype, for instance.

You have misread his point. He's not saying that users expect to see their contacts on their "first load". He's saying they expect to see their contacts after adding them.

Skype does fulfill that expectation, by keeping metadata about your contact list on their servers. Signal fulfills that expectation without keeping the metadata, by deferring to your phone's (local) contact system.

"Kings and dictators can also deploy new features faster than democracies, and they don't need to negotiate with anyone. Thus autocratic societies are simply better." This is literally how you and Moxie sound to everyone rational when you spout these opinions on centralization.

Nobody as far as I know, is arguing that a dictator ran software can't effect change faster. We are simply arguing that we would rather take a less effective system that has to compromise on certain aspects, over a unpalatable centralized solution. These arguments about 'We want the greatest amount of privacy for the largest number of people' are straight out of Plato's Republic.

In this case, the parent commenter would rather recommend that people use a less secure messenger so that he can sysadmin his own phone, and the Signal project has taken the eminently reasonable alternative position that getting cryptographic privacy into the hands of the vastly larger set of people who would never even imagine configuring an alternative app store takes precedence over satisfying the noisome demands of nerds on message boards.

You are of course welcome to use an alternate secure messenger for your own communications (I like Matrix just fine for what it is), or, for that matter, to stand up another instance of Signal, which is open source; there's a whole community of people who apparently do this.

The problem isn't simply that people disagree with Signal. Disagreement is fine (though recommending a "secure messenger" that isn't E2E by default is security/privacy malpractice, for whatever that's worth to you). The problem is when people can't metabolize disagreement and insist, repeatedly, over the span of years, that any disagreement must be evidence of malign intent.

You might be underestimating the number of rational people who don't think systems of government and instant messengers are usefully analogous.

Reply on news.ycombinator.com