matrix-doc VS Signal-iOS

Group voice and video calls are not E2EE, and use Jitsi, but this is expected to change with Native Group VoIP Signalling.

DMs in Matrix are always E2EE, and MSC3401: Native Group VoIP Signalling means there should be E2EE in group calls.

From my perspective, the really exciting thing about this that it works equally well in mobile web browsers as well as desktop web - clicking on a link on Mobile Safari should Do The Right Thing without having to install anything.

Moreover, because it's built on Matrix, MSC3401 (https://github.com/matrix-org/matrix-doc/blob/matthew/group-...) means that we'll finally have decentralised cascading video/voice conferences once the SFU (selective forwarding unit) component is added into the mix. So, for instance, users on the same homeserver will get their video feeds relayed locally with minimal latency... and then users on another remote homeserver will also get mixed locally with minimal latency, trunking the two together. If the link dies or one homeserver dies, the conference will keep going - i.e. precisely the same semantics as normal Matrix.

What metadata does Matrix protect? Encrypted state events still aren't a thing for example https://github.com/matrix-org/matrix-doc/pull/3414 This means that server admins know what groups a given account is a member of, private or not, and they also have a general idea of what the topic of said groups are, even if they're encrypted. This would be a problem for groups about sensitive personal medical issues, like a private HIV survivors or Alcoholics Anonymous group.

The protocol itself is flexible and can be changed through spec change proposals on their Github. They're currently working on implementing threads, and they recently implemented spaces, which functionally combine the concept of Discord servers and server folders. They can also be nested.

At the moment this only works with Jitsi. It will be implementet soon with MSC3401

Something we're trying to do about this on the Matrix side is MSC2716 (https://github.com/matrix-org/matrix-doc/blob/matthew/msc271...) - the ability to import archives of existing content into Matrix, and thus 'lock it open' and decentralise it for posterity: as long as one of the servers participating in that room stays alive (and the room is set up with infinite data retention, obviously) then the conversation will live on forever. (That MSC is also well worth a look for those interested in how Matrix works under the hood; MSC2716 was a surprisingly tricky problem to solve but it's basically finished now!).

Our first step will be to import all of Gitter's archives into Matrix - but we're then planning to add MSC2716 to all the existing Matrix bridges so that folks can use it to liberate chat history from Discord and Slack if desired, and avoid it getting paywalled/siloed/lost/held-hostage forever. We're also expecting to do USENET, mailing lists, forums, public IRC channels which have explicitly opted into logging... and generally archive as much possible in an open decentralised fashion, and ensure that gatekeepers can't lock up and blackhole info going forwards. After all, information longs to be free :)

by 'broken links' i guess you mean https://github.com/matrix-org/matrix-doc/issues/3628? it's a bug on the new spec website; we're working on it.

Fortunately, they did foresee this! The push notification only contains enough information to tell the phone that it should fetch the actual notification content from Signal's servers.

Here's a Signal dev talking about it on the Signal-Android GitHub: https://github.com/signalapp/Signal-Android/issues/12961#iss...

And similarly for Signal-iOS: https://github.com/signalapp/Signal-iOS/issues/962#issuecomm...

This was a nice, detailed read. I was happy to note this about employee compensation since paying them well is a good thing apart from their personal motivation to work on this (even at a comparatively lower pay than in other companies/projects):

> When benefits, HR services, taxes, recruiting, and salaries are included, this translates to around $19 million dollars per year.

> We are proud to pay people well. Our goal is to compensate our staff at as close to industry wages as possible within the boundaries of a nonprofit organization.

That said, I really dislike Signal for a few reasons. The first is what many people have already talked about very often — forcing to use a phone number to register. Since the SMS or call costs are quite high, Signal could adopt the iMessage approach to verification, which is having the user send an SMS to the service (this will cost the user some money depending on which country the SMS is sent to). This could be decided based on the country code so that the current SMS OTP model can coexist.

Signal is obstinate on a few aspects on user experience, more so on iOS/iPadOS. Firstly, it refuses to provide a data backup mechanism for iOS/iPadOS. If someone loses their devices, there is no way to restore older messages. Even setting up a new device requires the old device to be in physical proximity to transfer the data. Signal does integrate with CallKit (to act like a phone app) and with Apple’s notification services, but refuses to allow the user to backup the data with a password to encrypt it.

Secondly, I found this paragraph in this post to be disingenuous:

> Such practices are often accompanied by “growth hacking” and engagement maximization techniques that leverage dark patterns to keep people glued to feeds and notifications. While Signal is also free to use, we reject this kind of manipulation, focusing instead on creating a straightforward interpersonal communications app. We also reject business models that incentivize such practices.

Signal on iOS/iPadOS wants the user to enable notifications and to share contacts. If notifications are disallowed and if contacts upload is disallowed, it will pester every few days about it. One might think this is a silly mistake that Signal isn’t aware of. But it was reported some years ago and Signal responded that it will not fix it because it believes this is the only way. [1] Not even an option where this is a toggle for those who want no notifications or don’t want to share contacts (Signal does have a toggle for contact joining notifications).

Signal is also not that reliable in delivering messages in a timely manner compared to other apps (the GitHub repo has many repetitive issues on this topic over all these years).

Finally, since Signal has poorer UX in general, which isn’t an easy or cheap thing to handle, I use it only with less than a handful of people who I know and who use it.

I’d donate occasionally so that Signal can continue to exist, but I don’t feel like supporting it every month with all these issues, some of which look like Signal showing me the middle finger.

[1]: https://github.com/signalapp/Signal-iOS/issues/4590#issue-72...

I also use both extensively and I agree, Telegram has the best UX of any messaging app. The desktop app is QT I think so it feels very snappy compared to Signal's electron app.

Signal's iOS app is a native Swift/UIKit app and open source however, so I'd encourage you to report any bugs or issues you find: https://github.com/signalapp/Signal-iOS

Yeah I was able to find a few tasks that might be similar to your issue, though I don't know the exact details how the app behaves in your case, as in if you get error message or something like in this issue, or if it just closes on open like in this issue comment. Regardless, the generated crash report from the phone itself is probably good to add in your email to support (which is different from the debug log, just to specify in case you have been sending only that, here's a link for more details).

My bet is on "object-oriented obfuscation" and bloat caused by "modern" coding styles. Looking through the Signal source you linked to, I see tons of the former. There is an absolutely insane amount of boilerplate-looking code in here, for example: https://github.com/signalapp/Signal-iOS/tree/main/Signal/src...

I don't know Swift or the iOS UI API, but a lot of that code seems to be manually creating and positioning UI elements and otherwise implementing UI functionality that I'd expect the OS to handle. It's a lot of code that should really be data (arrays, structures).

To offer a huge contrast, a long time ago I wrote an MSNP chat client in pure Win32, and it was around 2kloc. The binary was 24KB. Of course it only supported contacts, presence, and text messaging, but I still can't see features like audio/video or the crypto that Signal has needing 100x more, especially if they use OS functionality for much of them.