Re: [Linphone-users] Why Android (Oreo) phones, are actually less reliable with TCP vs. UDP

[Brian J. Murrell]

On Sat, 2019-03-30 at 09:58 -0400, Greg Troxel wrote:
> 
> One is that pure-ish android (as shipped by google on pixels, and by
> lineageos, more or less) has some amounts of power saving.  Other
> vendors seem to add more power saving features, some of which seem a
> bit
> much, and on various lists I see various complaints about them.  So
> the
> notion that Juha does not see problems on "Android 7-9" and you do
> does
> not seem as inconsistent as it might.

Undoubtedly.  I fully understand that power-saving strategies vary by
device and manufacturer.  As a Huawei user, I have first-hand
experience of such aggressive strategies.

But I don't necessarily disagree with them.

As an example contrast, my Huawei device is the one which becomes
unreachable over wifi moments after the screen goes off, but is able to
wake up for FCM messages.  But my Nexus 5 with LineageOS 15 remains
pingable, for minutes (at least) after it goes to sleep.

> Some people really do not want to use FCM, partly because it's a
> non-Free network service, and partly because of concerns about
> tracking.

Sure.

> So having "use FCM" be the only reasonable approach is not ok.  (I
> realize some people want to use it.)  People that want to avoid FCM
> may
> end up with a bit more battery use, but that's for them an ok
> tradeoff,

Agreed.

> and as long as their phone reliably lasts between when they would
> have
> charged it, they probably think that's ok.

Also agreed.

> I certainly see your point about wanting to use the full-on FCM sleep
> method for extreme battery saving.  But when doing that, I think
> network
> use has to be integrated with the FCM wakeup notion a bit more
> deeply.

How so?  It seems to me to be working sufficiently.  When the phone
sleeps, wifi stops working *except* for FCM packets.  I imagine there
must be something in the wifi hardware that allows it to go into deep
power saving mode where only FCM (TCP port 5228, 5229 and 5230, AFAIU)
wakes it back up into power-gobbling mode.

But as soon as my phone gets that FCM message, it's wifi is alive and
well.

> There are some IETF specs about integrating push notifications, and
> my
> dim recollection is that there is a notion of shutting down the
> normal
> connections and then checking in and restoring the session layer
> state
> on wakeup.
>   https://datatracker.ietf.org/doc/draft-ietf-sipcore-sip-push/
> I recall sub-documents about FCM, APNS, and some non-proprietary
> websocket mechanism.

I certainly would not be averse to an open system for achieving the
same result, to be sure.

Cheers,
b.

signature.asc
Description: This is a digitally signed message part