Some people seem to misunderstand this thread as an invitation to evangelize XMPP or Matrix in my mentions.

That isn't an invitation at all.

I will instead invite you to fuck off into the sun.

@soatok

I do want to know more about how signal solves (or doesn't) a specific problem, and i'm not sure your blog answers that. (At least i don't remember that it does)

Spcifically, what happens between 'private contact discovery' and sealed sender taking effect? Afaik, the envelope protects only after keys have been exchanged, but requesting keys would expose the social graph to a compromised signal-server instance. Do you know more or is this just not protected, only 'live' metadata?

@soatok Urgh, Telegram’s the worst. Unfortunately I have to use it for one specific purpose, because reasons.

Is there any comparison of Signal with Threema?

I have been using Threema for longer than Signal exists, and I like it. The UI is really good, in particular the visualisation of the trust relationship with the other (unknown / known contact / checked public key => 🔴 / 🟡 🟡 / 🟢 🟢 🟢). I keep hearing Threema’s cryptography is rather good, but a comparison could be interesting.

@matildalove @soatok Part of the challenge is that very few people actually get placed into the situations where they can genuinely appreciate security-oriented design features. In other words, there are a lot of posers that want to roleplay as security experts.

Another is that it's really easy to think your shit is secure, when in fact you haven't put enough thought and effort and testing and monitoring to really support that belief.

Also the gatekeeping culture surrounding cryptography is absolutely insane. It is *by far* my least favorite community to try to interact with, by a long shot.

I mean, when a certain prominent Haskeller tried to reinvent a message authentication code, I took the time to actually develop an attack to break it. He didn't understand what I had done as an attack, at first, but he came around and migrated to HMAC. In the end, that story was a success.

On the other hand, if you do actually do something that is geniunely interesting or useful, everybody will tell you to stop doing cryptography, without actually clearly articulating an issue with your idea.

@darkcat09

A centralized service can go down, federated one cannot.

Tell that to the pawb.social outages.

So I still prefer self-hostable solutions over Signal, as they also have reliable cryptography.

No, they fucking don't.

@david_chisnall @soatok
> Signal or Google being forced to push out malicious updates to the client

Especially since Google has been working *hard* to enable that threat model.

It used to be that apps were end-to-end signed by the developer, and the devices would refuse updates signed by the wrong key.
Then Google offered a "convenience" feature where they held the signing key.
Which is now mandatory. Google can, at any point in time, just decide to ship an update to any app.

@soatok @david_chisnall I don't think people are failing to grasp how that is hard.

I think the thing that often times gets missed here the context in which it is being discussed, which is a alternative to SMS and not a generalized messaging system. In which it is a really good and secure one.

But aye, lots of the security complaints about Signal are bullocks.

@deetwenty @soatok

I completely agree. I recommend Signal because the alternatives are worse. That doesn’t mean I’m happy with Signal. There’s a lot I’d like them to improve. But improving Signal is a lot easier than building something new and making everyone switch or getting one of the other alternatives to the same end goal.