Email or username:

Password:

Forgot your password?
Delta's posts Post Back to profile
Top-level
Lyyn ☮️

@fell @delta @rakoo I see people blame optimizers for forbidding persistent connections, but I prefer thinking that it is a conceptual problem with mobile environments. In the end to keep the connection you need some program to be loaded, and it can consume RAM and run at random moments. It should be possible to completely kill a program and it still should work correctly.

14 May at 10:45 | Wall-to-wall | Open on mastodon.ml
9 comments
Delta Chat

@lyyn @fell @rakoo surely, push notifications are a complex thing only arising because mobile OSes prefer to prevent apps from just sitting there, listening for network activity. dontkillmyapp.com/ exists for good reasons.

14 May at 12:01 | Open on chaos.social
Felix Urbasik

@delta @lyyn @rakoo I think the real reason is because (Android) smartphones want to suspend the actual CPU whenever possible and receive notifications via the less power hungry baseband modem. AFAIK it's capable of maintaining simple TCP connections on a very low level, which is what Google's FCM service is.

14 May at 14:08 | Open on ma.fellr.net
rakoo
@fell

Read somewhere a loooong time ago that the reason for that is that some applications were doing stupid stuff like sending packets all the time, essentially having a constant use of the network, just to check for new content. When all applications do that, each on its own clock, the battery doesn't last very long.

Which is a shame because a proper application without FCM and proper implementation is virtually invisible. Keeping a TCP socket open costs nothing, even from an app. But it's always the same: if the consequence is that the battery is bad, then the one that is blamed is Google/Apple, not the app maker.
@delta @lyyn
@fell

Read somewhere a loooong time ago that the reason for that is that some applications were doing stupid stuff like sending packets all the time, essentially having a constant use of the network, just to check for new content. When all applications do that, each on its own clock, the battery doesn't last very long.
14 May at 14:30 | Open on blah.rako.space
Lyyn ☮️

@rakoo @delta @fell I used Signal on websocket without FCM and had a feeling that battery life is not perfect. I believe Signal spent some effort not to do it the stupid way.

14 May at 14:33 | Open on mastodon.ml
Delta Chat

@lyyn @rakoo @fell in our measurements delta manages to consume less battery when there are no gplay services as it can do efficient idling. Google fcm and play have their own overheads even if their purpose it to prevent apps from consuming too much battery. With the upcoming FCM Push Service integration for Chatmail users we might redo these measurements but there is a wide variety of setups on Android so not so easy.

14 May at 14:40 | Open on chaos.social
Lyyn ☮️

@delta @rakoo @fell Hmm, if you are integrating FCM into Chatmail do you have plans to add UnifiedPush support? I saw a discussion thread on your forums, but it was closed 3 years ago.

14 May at 14:44 | Open on mastodon.ml
rakoo replied to Lyyn
@lyyn

Not a delta dev, just my opinion: UnifiedPush is useless here. Either you want to integrate with FCM because reasons, or you use the native IMAP IDLE that doesn't consume much. There is no need for an intermediary because there is no need to make it pluggable.

UnifiedPush is useful for apps that don't use a protocol with a proper push mechanism even though it's important for the app to work. If anything it's a confirmation that HTTP is crap and should not be used *that* much.
@delta @fell
@lyyn

Not a delta dev, just my opinion: UnifiedPush is useless here. Either you want to integrate with FCM because reasons, or you use the native IMAP IDLE that doesn't consume much. There is no need for an intermediary because there is no need to make it pluggable.
14 May at 14:52 | Open on blah.rako.space
[DATA EXPUNGED]
rakoo replied to DELETED
@lyyn

There already is a standard though for sending messages, that doesn't consume much and can interleave multiple payloads, it is XMPP and is specified by the IETF. In fact FCM allowed XMPP at some point. I'd agree with you that XMPP should replace all those reinventions and both Google and Apple should use it, but alas.

It totally makes sense to use UnifiedPush instead of websocket though, because websockets are a hack because using HTTP here is silly, and it requires specific machinery. It's easier to rely on another party. Deltachat isn't the same: there is something that works natively, that should be the preferred way of doing notifs
@delta @fell @grishka
@lyyn

There already is a standard though for sending messages, that doesn't consume much and can interleave multiple payloads, it is XMPP and is specified by the IETF. In fact FCM allowed XMPP at some point. I'd agree with you that XMPP should replace all those reinventions and both Google and Apple should use it, but alas.
14 May at 16:13 | Open on blah.rako.space
Григорий Клюшников replied to DELETED

Lyyn ☮️🦄 #2024-02-16, Mastodon doesn't use UnifiedPush, it uses a WebPush -> FCM relay. There is currently no way to receive push notifications without Google services or at least MicroG.

14 May at 22:08
Go Up