The Signal Server repository hasn’t been updated since April 2020. There are a bunch of links about this here but I found this thread the most interesting.
To me, this is unforgivable behaviour. Signal always positioned themselves as “open source”, and the Server itself is under the best license for server software (AGPLv3 – which raises questions about the legality of this situation).
Signal’s whole approach to open source has constantly been underwhelming to say the least. Their budget-Apple attitude (secrecy, i.e. “we can never engage the community directly”, “we will never merge/accept PRs”, etc) has lead to its logical conclusion here, I guess. I have been somewhat of a “Signal apologist” thus far (I almost always defend them & I think a lot of criticism they get it very unfair) but yeah I’m over Signal now.
A few years ago (2017?) I decided I would move messenger apps. The aim (and what I’ve achieved) was all my messaging going through a secure, private app.
In 2017, Signal really was the only option. Element (Riot, back then) was really bad and didn’t feature e2ee (which only got enabled by default last year!). XMPP was and remains difficult to use (not even many people here use it, how could I expect “normal people” to use it?)
I made the choice to use Signal, and I don’t regret it. I only regret that it has taken until now that we are starting to see a glimmer of a real competitor, in the form of Matrix. But a really competitor to Whatsapp and the like, back in 2017, just didn’t exist outside of Signal.
I don’t quite get why you think XMPP is harder to use than Matrix. The only way this seems true is if you use the main matrix.org instance and then you are pretty much back at a centralized service based in a five eyes country (UK).
It’s not about instances, they’re pretty much equal in that regard. There are two main issues with XMPP:
I disagree on the default client idea, especially if it is such a badly done web-based one as element/riot. In the end clients are always platform specific, and there are easy to find “best” XMPP clients for each platform. At most it is a branding/marketing problem (see Snikket.org for that).
As for the other two points: that is both false and outdated. e2ee has been supported and the default in XMPP for longer than in Matrix and message history (as much as the e2ee double ratchet algorithm used both in OMEMO and OLM permits) is working perfectly in XMPP across clients if the server has MAM enabled (pretty much all have).
You say you disagree with the default clients idea, but why?
I don’t know why you’re so dismissive of this issue. I feel like you’re framing me as if I’m anti-XMPP when that isn’t the case; on the contrary I use XMPP and am a Prosody server admin. The reality of the situation though, like I’ve said above, is that next to nobody uses XMPP, even in tech communities. At this point “branding/marketing” could end up being the be-all and end-all of the entire protocol.
You’ve misinterpreted my comment. I am very well aware XMPP has and has had e2ee support, the issue is that XMPP clients never have this switched on by default, in my experience (which was testing every XMPP iOS client there is, the platform most my friends use).
The situation on iOS for XMPP is uniquely bad (but actually Siskin and Monal are improving a lot lately). It simply is unfair to look at only one tiny and for most people irrelevant (and uniquely bad) platform and extrapolate from that. On all other platforms XMPP works great and most clients have e2ee enabled by default for 1:1 chats (where it makes the most sense).
As for the other topic, sorry I didn’t want to sound so confrontational, but the same argument comes up all the time inside and outside of the XMPP ecosystem and I think it is simply false. There are other problems why XMPP isn’t adopted. Network adoption is driven by network effects. People invite other people to the network and when doing that they typically also recommend a client (& server). The case of a lone person looking for a new messaging system without any network is the rare exception and one that only comes up in the bubble where this discussion usually takes place.
Well there was Wire, which offered e2e encryption, an open protocol and opensource clients and backend, it has been audited, and it was based in Swiss which is times better than the US. I tried to move a lot of people there, but luckily I failed, considering it has been bought by an advertisement company recently
Wire was pretty good, true. I used it a bit, but chose Signal because Wire (similarly to Matrix, for now) doesn’t encrypt any/most metadata, whereas Signal encrypts everything and always has.
And like you said, it’s since been sold to an advertising company. Not sure if that’d even be possible with Signal since it’s owned by a non-profit (admittedly not always the case, I guess it could have been possible when they were still OWS).
In both cases, their centralised nature means changing ownership can be devastating (like in the case of Wire). This is why I believe Matrix is the future. Its community is much healthier and active in the development of the ecosystem (3rd party clients, bridges, they actually accept PRs, etc…)
This is not exactly true. Encrypting metadata is most times impossible due to the server needing to know who to deliver messages to (at the very least). “Sealed sender” is now a thing (though i don’t know how strong a protection that is), but to my knowledge Signal continues to aggressively expose users’ phone numbers both to the server (in a hashed formed, for contact discovery) and to other users in public chatrooms. Please correct me if wrong.
A non-profit doesn’t mean you need to do good. Also, it can turn into a for-profit over the years. It’s in fact a conscious strategy of startups in the field of “sharing economy” (remember couchsurfing?)
Matrix is one among others, but i’m not convinced a single solution is going to be the best:
They all have strong arguments going for/against them. I believe interoperability is the only way to go. These network are doing mostly the same thing, and there’s no reason we can’t talk across networks.
Which brings me to the fact matrix folks really don’t seem to care about interoperability though i hope i’m wrong about this.
I have a lot of thoughts about this but don’t really have the time to reply.
All I’ll say is that I hope you’re following Element’s progress with Dendrite closely. I host my own Dendrite server and it is much more reasonable in terms of resource usage versus Synapse, and it hasn’t even had any resource optimisation features implemented yet.
While Dendrite is better in many ways, AFAIK it does not solve the fundamental architectural problem of immutable and permanent history room metadata. As a result of that, database storage use is growing indefinitely (easily into the hundreds of gigabytes) and there is no real solution to that anywhere in sight. In addition I think it also is a massive privacy issue, as this immutable and permanent history room state data is synchronized across any server that has a member joining a chat. Yes I am aware that this is a “feature” of matrix, but IMHO a really bad one and resilient federated rooms can also be implemented in different, less over-engineered ways.
This is terrible.
Matrix evolved evolved in a very messy way, starting without encryption and hacking it in later on, and now it’s even trying to become P2P. I expect more serious privacy-breaching “features” to come out over time.
Not really, that was a feature that was there from the very beginning and Matrix also openly advertised this. The problem mainly comes from people projecting their wishes onto them and the Matrix team (for commercial interests/ego I guess) not vehemently denying that privacy is mostly an afterthought in the system’s design.
I’m keeping an eye on Dendrite. I’m not convinced go is the best language for server software, as it suffers many same pain points as Python (eg. GC pauses), but it looks like a neat progress. In fact i’m going to try dendrite very soon when i have some time.
Element on the other hand i would just put in the dumpster because it’s full of everything that’s wrong with web applications. 9MB initial loading just for a simple chat application, seriously? Several seconds of latency just to switch chatrooms? Seriously it’s 2021 folks, how can anyone be happy with such mediocrity and then complain why noone is using Element…
Just found gomuks which appears to be a lot better for desktop/laptops (not mobile). I will try it out and see…
So i just tried gomuks and it’s a pleasure to use! Room switching is instant (compared to 5-15s on Element) and it took just a few seconds to compile. Only downside is it was designed for dark theme so contrast is really bad on light background.
Element the client is garbage, I was talking about Element the organisation formally known as New Vector, who develop and maintain the Dendrite homeserver
thanks i had no clue they were renamed
FluffyChat is a decent alternative client (with E2EE support). If you don’t need e2ee there’s actually a healthy number of clients, and some of them do seem to have it on their roadmap
https://matrix.org/clients/
Point taken on server implementations though
FluffyChat is not an option because it doesn’t support proxies including Tor. If you’re using fluffychat please open an issue there for integrated tor support like Conversations/Gajim does in the Jabber/XMPP world :)
That pretty much sums it up. Matrix isn’t bad, but basically over-hyped and reinvents the wheel for most stuff.
As for sealed-sender in Signal: That is in theory a good idea (and should be implemented in XMPP at some point), but in a walled garden with a single server it is snake-oil as the central server can still easily correlate sender based on other metadata.
Not only sold, I used to report bugs to Wire by e-mail and GitHub before of the change.
One day, they just sent me an automated message in which they said they would not going to provide support to the personal edition at all during a time because of the lack of staff while providing support to the business edition.
It passed more than a year and was maintained, I don’t know today but I expect the same.
Edited: I don’t know why I put Signal instead of Wire jajajajajajajaja.
Wire looked nice, but I stopped using it after they persistently dragged their feet on federation.
Git discussion
Once something with federation gains popularity, the discussion may be over, as we won’t have to talk about jumping ship every year. I’m not sure it’s doable yet, but I’m sure that once it takes hold it’ll last, just like email.