1. 19
  1. 20

    This post contains the same 3 complaints you’ll routinely see about Signal: use of phone numbers to identify contacts, use of Google Cloud Messaging, and lack of federation. Granted these are all valid complaints, but are essentially true of any of the popular “secure” messengers which support asynchronous messaging between participants who may-or-may-not be online (Emphasis on popular here: there are niche messengers who solve one or more of the afforementioned problems, often to the detriment of user experience)

    It then goes on to… prescribe nothing? Instead we get this:

    The big question now, as also said by @shiromarieke on Twitter, is what post-Signal tool we want to use. I don’t know the answer to that question yet, but I will lay out my minimum requirements of such a piece of software here.

    The rest of the piece takes on a “I want a pony” air of wishful thinking to it. The oddly ironic part of it is this sort of longing for better encrypted messaging is exactly what people were doing in the decades prior to Signal, despite him describing his train of thought as “post-Signal”. Those who forget the past are doomed to repeat it, I guess.

    Unfortunately just wishing for something doesn’t make it true, and in the meantime Signal is real and a best-in-class tool for secure messaging use cases.

    Since the author didn’t, I’ll go ahead and give a shout out to Matrix here:


    I think it provides much of what the author wants. But I probably wouldn’t recommend it over Signal, yet: it’s a work-in-progress and doesn’t necessarily cover all of the Signal use cases yet.

    All that said: I found this to be a fairly substance-free post: tl;dr: Signal isn’t perfect? Cry me a river. Perfect is the enemy of good

    1. 9

      Yeah, I mean, this is somebody who trains journalists in secure comms? So what does that training consist of now? “I recommend you not be a journalist.”?

      1. 1

        I don’t understand this comment. You say that the post contains three valid complaints, but he shouldn’t complain, because it’s still better than some other tool. With WhatsApp using Signal’s encryption now, I can’t see much difference between the too: You are still in a walled garden, with your metadata going through google. I don’t get why you ridicule these concerns.

        I second the matrix recommendation, though. It’s everything Signal promised to be.

        1. 1

          I didn’t write this comment, but I think I understand it:

          Signal is probably the best we have right now in “production quality” state, so it’s the best option we have, but we could do better. In that case it might be better to support the “not yet production quality” stuff.

          Or super short and offensively said: Shut up and hack! (as the OpenBSD people would say)

          Mainly, because most people are aware of it anyway, but that doesn’t make it better and not recommending the best available option seems kind of a strange move - at least to me.

          Like I said, my understanding. I don’t know for sure, if that’s what bascule actually meant. But maybe hearing it through someone else’s words helps understanding. :)

      2. 8

        I use signal as a compromise. Generally if I want private communication I would send a GPG encrypted email.

        Problem is I don’t trust my phone. It can be exploited/rooted easily as I’m on the vendors mercy to have a properly patched Android device and even if I self rooted the phone there’s still the base band having much grander access.

        Now Signal solves 2 problems for me.

        1. I don’t trust my phone so I won’t put my GPG material on it hence I can’t decrypt or encrypt email from my phone.
        2. It’s easier to convince my contacts to just install Signal

        The threat model here is pretty obvious - I already consider my phone owned if targeted by a serious adversary, so anything sent or received from it is not relevant. Yet there is no need for anything I send/receive being clear text. It’s just a small improvement for regular communications that might prevent my mobile provider, kid next door snooping over messages.

        1. 4

          Fair analysis. I like that. Similar to how I use them. Shit, I haven’t even upgraded since the Snowden leaks despite being on the Next plan. I just got off it lol. Each model upgrade of leading phones seems to make surveillance easier up to making batteries hard to remove, extra camera, mic’s that might be on… it’s ridiculous. Even JackPair will require a smartphone with an off switch for mic (sound) and camera (lip-reading).

          Current scheme is air gapped machine for GPG with me just not trusting anything on a smartphone or regular PC. I don’t care what OS it’s running if it’s Intel/AMD hardware.

          1. 2

            I don’t trust my phone so I won’t put my GPG material on it hence I can’t decrypt or encrypt email from my phone.

            You can use a smartcard with NFC, like a Yubikey, to perform GPG operations using OpenKeychain on your phone without exposing your private key to Android. It’s a great system in my opinion.

            1. 4

              Both black hats and nation states have been very successful attacking Android phones. You’d need two partitions, untrusted and sensitive, on a separation kernel at a minimum with protection of all chips' firmware. Most solutions just do former. Nobody does the latter that I know of with it getting harder over time with smartphones containing a boatload of chips.

              So, you shouldnt trust a smart phone for your secrets unless your theat profile is lower than average malware. Snooping, non-technical parents and such.

              1. 3

                I don’t trust my phone. If I decrypt or encrypt anything on my phone then it means it either went from crypto -> plain-text or the other way around. If my phone sees plain-text then an adversary can also see it.

            2. 6

              There are however, multiple issues with Signal, namely:

              Lack of federation Dependency on Google Cloud Messaging Your contact list is not private The RedPhone server is not open-source

              I too would have liked XMPP to take over the world but nobody made an easy-to-use very secure client. All sorts of collaboration tools (ala Google Docs, iCloud documents and screen sharing) could have tunnelled through extensions of XMPP.

              But these complaints against Signal make as much sense as “I don’t like Signal because rhino babies are cute” or “The moon is 384000 km away, hence I won’t use Signal.”

              1. 8

                I’m glad XMPP didn’t take over. It’s horrendously complex and inefficient. PSYC people had a nice page on that:


                PSYC protocol was a contender with better simplicity & efficiency. See “old” link for how simple a minimal implementation is. They’re redoing it for better privacy right now.


                1. 4

                  Seconded. I tried using a private xmpp server. It’s a minefield.

                  1. 3

                    As someone who set up a couple of XMPP servers, I agree on the statement, that XMPP isn’t the best protocol, but I am not sure by what you mean with “it’s a minefield”. It takes a bit of effort in the beginning, but it appears to not be harder than various other servers whose main purpose is communication (IRC, SMTP/email, etc.).

                    Was it on the protocol or the implementation side? Cause on the implementation side I have seen huge differences.

                  2. 2

                    PSYC also didn’t take over (I don’t know it enough to comment on whether that’s good or bad), but I have high hopes in Matrix, which I only really read into after it was mentioned here. I like the really pragmatic approach.

                    That said I never found something really negative about PSYC, still it just never picked up. It appears interesting though and I like to link to that page too as a reference for what’s bad about XMPP.

                  3. 4

                    XMPP is horrible to configure & set up. The fact that you have a ton of extensions that some servers have and some don’t is not helping. There are so many off the record client implementations and guess how many went through a solid security audit?

                    I used XMPP in the old days when everyone did. It was decent (though not better than existing commercial solutions). I tried it again a year or so again and running my own XMPP prosody server. I stopped even starting up my client - had like 3 contacts.

                    I am seriously considering just whacking that server as it’s an unused liability.

                    1. [Comment removed by author]

                      1. 7

                        People who say “I would totally pay for this” >>> People who would totally pay for this.

                    2. 5

                      A few of us have been using http://matrix.org/ which is soon to support end to end encryption for group chats, is fully open source, and does not use a phone number for an ID.

                      I use the weechat plugin and the android client. Seems to work nicely, but not tried the crypto yet. Last I heard, the crypto was under audit.

                      A few of us hang in the ‘OpenBSD’ channel there.

                      1. 4
                        1. 3

                          Oh wow, so Moxie stated over three years ago what they’d need to avoid Play services, and five months ago what they’d need to get a WebSocket-only non-GCM version, and so far there’s been no pull requests just a bunch of blogs posts complaining that someone else hasn’t done the work yet. Open source at its best.

                          1. 1

                            True. From my limited experience that isn’t too different from Closed Source though. The only difference is that it’s less noisy in closed source environments, or kept completely internal.

                            My limited experience though.

                            1. 1

                              But with open source you are still able to fork, fix the issues, put your own servers online and release your own service.

                        2. 2

                          foxy moxies arguments are raising the question why he’s so concerned about people not using the one true signal app with their servers. either the server software is seriously bad and breaks with unexpected input from other clients or their app is doing / will be doing something funny.

                          either you want to make the world better and provide the servers without restrictions or stfu and be whatsapp.

                          i can to some degree follow the argument about the name “signal”.

                          1. 3

                            I don’t think he’s worried that the server would break, but that people using the official Signal app would experience issues communicating with people who use LibreSignal and think Signal itself is broken. Such as messages being slow to be delivered when the WebSockets code doesn’t wake the recipient’s phone up, or when calling someone completely fails because receiving calls doesn’t work without GCM.

                            1. 1

                              people using libresignal can be expected to tell their peers that it’s their client which is responsible for the faults. what about a message notifing peers about possible problems with a 3rd party client? should be possible.

                              this feels like icq changing their protocol all the time. we are making progress, do we?

                              1. 1

                                people using libresignal can be expected to tell their peers that it’s their client which is responsible for the faults

                                That’s one hell of a ‘citation needed’. I’d expect the opposite to be true.

                                1. 1

                                  There is no citation.

                                  Maybe I still have too much trust in common sense. My rationale was that libresignal isn’t installed “by accident” but by decision to use it (instead of the much more visible official signal client). People doing that usually can’t use the official client because of other decisions they’ve made, most of the time deliberate decisions based on not wanting to use gapps/android and have to do explaining all the time as they use a non conforming setup.

                                  Edit: Fixed my english a bit.