Moxie is open to someone writing an alternative to GCM, but nobody has stepped up and do it. It's much easier to complain about GCM, then to replace it.
Signal's use of GCM does NOT reveal who you are sending signals to, or what is in the message.
Signal's use of GCM does NOT reveal who you are sending signals to, or what is in the message.
I might be ignorant about GCM. My understanding is that first you need to register with GCM and have the connection stay online permanent. That requires Google's Apps, which open the system to Google's Access, which clearly is not acceptable. Secondly, Google knows that way always where I am. I do not want a permanent connection to Google servers. Also, how does GCM know where to deliver my messages when it can't know who I'm contacting? I'm not saying that that is an impossible task, but to my knowledge Google isn't reliably (reliable in the sense of keeping my privacy even if Google were to become the attacker) solving it. [0]
Each of the three reasons given above are enough to not use GCM.
That's not quite true, LibreSignal [1] was great while it lasted but Moxie killed it [2].
[1] https://github.com/LibreSignal/LibreSignal
[2] https://github.com/LibreSignal/LibreSignal/issues/37#issueco...
Don't confuse the issue with LibreSignal. Problem is Moxie is worried about the security of random 3rd parties publishing signal clients. Doubly so if published by f-droid (which doesn't' sign package with the developers key). Additionally he wasn't want these 3rd party client that may or may not be secure using the whisper systems infrastructure.
So it seems that signal could get a GCM replacement, if someone writes one.
https://github.com/LibreSignal/LibreSignal/issues/37#issueco...
I urge people to read through these comments in the github issue yourselves versus believe the commenter here at face value.
> If you think running servers is difficult and expensive (you're right), ask yourself why you feel entitled for us to run them for your product.
Moxie knowingly and willingly runs a free service. Why should he feel entitled to tell people how they interact with his service? What's the point of publishing your source code under the GPL if no one can do anything with it? Keep in mind the phone server still isn't open source either, so the client source on its own is about as useful as a literal brick for phone calls.
1) he doesn't want the signal community to fracture with incompatibilities (like say XMPP)
2) he doesn't want the security of signal users compromised with less secure clients. Thus he's against SMS since it leaks metadata like crazy to 100s or 1000s of entities.
3) he doesn't particularly trust f-droid to distribute clients since they use a central key (an attractive target) instead of the developers key.
4) he wants to be able to quickly iterate improving the clients and server in sync, on his own schedule.
So generally he wants to target a large number of users with a very secure client and doesn't want to sponsor anyone that takes his source with free server infrastructure.
So he believes in audits, security, and opensource. He even will allow a GCM replacement, as long as it only runs when GCM is not available. He expects the battery life and user experience to be terrible. The main problem seems to be that google negotiates with cellular providers to not time out connections to GCM, but they silently do that (without an RST) to other connections. Thus any GCM replacement will have to spend much more power/battery maintaining a connection. Moxie specifically mentioned that even 50 seconds was too long and that some providers time out connections even quicker than that.