Hacker News Reader

Comment by COGlory | Hacker News Reader

Signal always left a bad taste in my mouth. It's fundamentally built on the requirement that to have conversations with any trusted individual, I need to, for no obvious reason, also incorporate and trust a 3rd party individual, who won't even use their real name, and did some shady crypto currency integration.

Matrix is so so much better in that regard. I can communicate without any 3rd party needing to be trusted. The areas where I need a "3rd party" (the server and client code) are effectively diluted by the fact that thousands of eyes pore over it. Rather then trust in a single person (Marlinspike), instead I just have to trust the person I'm actually talking to.

And the Matrix team have done nothing to make me question their motives.

View on HN

I am a Matrix user -- I run a homeserver primarily for bridging my other services into one chat application -- and I am a Signal user.

I have found that Signal strikes a good enough balance with E2EE that I can easily recommend Signal to anyone -- I do not find the same to be true of Matrix. While I believe Matrix to be a better choice for freedom and privacy I do not see it as a good option for non-technical users. Even for technical users I feel it can be very confusing to use unless you are already familiar with the Matrix ecosystem, jargon, etc. With Matrix, the overhead to manage sessions and your security is much higher than Signal. Sure, it's unfortunate that you have to place some trust in Signal, but it's far less confusing than having to manage your own sessions.

In my subjective and probably biased experience, Signal feels as easy as iMessage but still provides greater privacy than most other platforms (not Matrix, obviously).

I really like Matrix for myself and for dev communities. But I have a hard time seeing a future where Matrix becomes a mainstream personal messenger until the experience feels a bit more hands-off for non-tech users.

100% agree. Matrix is great and while it took a bit to setup everything I liked the direction it's going.

Started using Beeper which is built on top of the Matrix protocol. Even with this nice abstraction app, there were several steps to set up and then connect to the various bridges they have. They're still very much a beta product.

Can these problems be solved with a good UX developer in Matrix? Or is it more fundamental to the platform? (I havent had the chance to try it yet)

> with Matrix, the overhead to manage sessions and your security is much higher than Signal.

Isn't Signal just as tricky to manage if you want to have multiple sessions?

No, with Signal on desktop or iPad it automatically opens a QR code for you to scan and once you scan it your new device gets linked.

With Matrix, when you sign in on a new device it shows a message the previous device to verify. Sounds about the same

Yeah, I guess that's true. I suppose it's the way Matrix gives you control to manage multiple sessions that gets confusing. I can understand why it's good to give users that control -- but for non-technical users it gets confusing quick in my opinion.

I agree, but I expect Matrix to catch up in the long run. There is nothing in the Matrix protocol that fundamentally blocks it from having a great UX. The client apps are just not as good as Signal's yet.

The same argument can be made for XMPP and it's already standardized by the IETF.

I'm in exactly the same boat as you. Signal for random friends and family, Matrix for anyone who can handle a little bit of tech. No one else in my Matrix friends group chat works in IT, though, so it's not that bad.

"I won't trust an individual who won't even use a real name"

That's the hard privacy-killing argument of corporations. And only cultural change can help it.

Funny thing is you can't use Signal without revealing your phone number.

So much for privacy.

...and they tell all your contacts that you've created an account. super bad.

anyway, I hear they're working on it.

Not your contacts, but everyone [already using Signal] who has you as a contact.

To remove that functionality would be equivalent to removing the ability to look up if a Signal message can be sent to a particular number. At least as long as phone numbers are used as the sole identifier for Signal accounts—another decision that has been debated into the ground, of course, but was done for understandable reasons.

yes, thanks for the correction.

that is mostly a UX problem, not a technical problem: people are surprised that signal tells others that they created an account, and this isn't a pleasant surprise IME. doubly so (more, actually) if you've got a stalker who has your phone number.

I've even had people start arguing with me as if I'd created the new conversation on their Signal list.

That would also remove the ability to easily transfer your existing social graph, which would kill adoption. You don't actually have any other choice if you target for mass adoption.

You're confusing privacy with anonymity.

    Privacy != Anonymity

If you live in a village where everybody knows each other what happens inside your house can be private even if everybody knows you live there.

The other way around you could be at a nightclub in a big city where nobody knows you and do the most i timate stuff publicly (but protected by anonymity).

Of course if you have privacy and anonymity what you do will have even stronger protections, which could be especially useful in the internet, where the biggest part of our conversations is not ephemeral but persists. But to defend Signal here, this is actually something you could choose with their disappearing messages.

Anonymity massively helps privacy. Legislation in my country could force the company to spy on me and break encryption, technical challenges completely aside.

a phone number is a very private thing so your explanation is unwarranted. you can be concerned about giving away your phone number without having concerns for anon

I would guess that for many people their mobile phone number is a identifier more stable and unique than even their full name. In some jurisdictions it's hard to get one without presenting an ID.

except as a spam impedence (not prevension) requiring phone numbers is a legitimate security and privacy risk, especially for people who think it is a stable and unique identifier

You don't have to trust a 3rd party. Just compare your security codes IRL or through another trusted channel.

Without doing that, you can never be sure if there is a MITM attack. Not with Signal, not with Matrix, not with WhatsApp.

EDIT: MITM may not be the best description, let's say: "you cannot be sure, that you talk to the correct person."

The problem with Matrix (compared to Signal) is the huge amount of unencrypted metadata your average Matrix conversation will generate.

I'm not versed well enough in network theory to decide if that's a consequence of federation or if it's just an oversight. I'm also not enjoying the wasted bandwidth of base64 encoding binary data in JSON, the encryption is inherently wasteful to make it easier to use for web developers.

I do prefer the ideas and concepts of the Matrix ecosystem, but Signal does privacy better in pretty much every way except for them requiring a phone number.

> huge amount of unencrypted metadata your average Matrix conversation will generate

Do you have more details on this? Matrix being distributed of course will not "tumble" everything through a single service, sure, is that what you mean or other issues?

Matrix servers store the conversation history for their local users. The message contents (for e2ee rooms) are e2e-encrypted. The metadata (which user sent it where and when and to who) is not currently encrypted or obfuscated. We're working on addressing this in the context of P2P matrix (as per https://arewep2pyet.com).

I dont really consider it too much of a problem but there are unique identifiers in plaintext for (almost?) every message to make federation and E2EE work. If an 3ncrypted, federated room receives a message, its participants need to know what key to use for decryption and in some cases what specific device ID was used to request the right keys.

What I find particularly strange is that my messages also contain the event identifier a message is a reply to outside the E2EE envelope, as well as the message type for example. If you can get your hands on a database, you can find reply messages and see the messages part of threads without having to decrypt anything. I'm not entirely sure why the server would need this information, perhaps for making notifications work efficiently?

You can see this for yourself if you enable dev mode in Element, hold/right-click a message and hit "show source". You can also compare that to the result of "show deceypted source".

All of this metadata is being transmitted over HTTPS, of course, unless you're using some local devices setup. It's not literally plaintext but the envelope stored on the server contains much more message details than Signal's.

There is nothing shady about the integration of MobileCoin with the Signal client.

You may not like MobileCoin (or perhaps not like cryptocurrencies in general) but to call it "shady" is inaccurate.

It was pre-mined. That is the very definition of shady.

And the integration was built-in secret (no commits on Github for over a year) to maximize the profit from pre-mining.

Reply on news.ycombinator.com