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

[Brian J. Murrell]

On Sat, 2019-03-30 at 15:30 +0200, Juha Heinanen wrote:
> 
> I'm not sure what you mean "pingable",

Can you ping it?

$ ping <hostname|ip-address>

> but it responds to SIP OPTIONS
> request when the screen is off

So the network stack is probably not sleeping the way it does in my
phone.

> If foreground process is not running, baresip does not respond to any
> SIP request.

Of course not.  But does it remain pingable (see above) for long after
the phone's screen goes off?

> But as I mentioned, based on my experience with Android
> 7-9 devices, foreground process stays alive until phone goes to power
> saving mode.

Sure, but does the network stack stay alive?  That's the issue here.

> Mostly I have been using mobile data in my tests, but last night I
> had
> an Android 7 device on and both wifi and baresip were still alive in
> the
> morning.

So, wifi being turned off when the phone sleeps does not mean that it's
not on when you wake the device up, such that you have to turn it back
on manually.  I'm referring to the phone putting wifi to sleep when it
sleeps and waking it up when it wakes up.

> Sure, if wifi is on, battery does not last as long as if wifi is
> off.  I
> just started baresip over wifi on an Android 7 device tha has 80% of
> battery left,  I'll check after a few hours how much of the juice is
> gone.

And see if you can still ping the device once it is sleeping a good
while.

> Depends on what else the phone is doing.  If nothing else, then
> perhaps
> 15 % of so.  I can check tomorrow morning after I have charged the
> battery fully.

15% is a lot.  You really want in the small single digits.

> None of the three phones that I have has turned off wifi
> automatically
> when baresip service has been running.

And neither should it.  But it shouldn't circumvent the phone's battery
saving desire to disable wifi either, by forcing the phone to keep it
on.  Assuming you are also not anti-push, because without push you have
to leave wifi on to receive calls.

> As long as you cannot run your own push notification server, you can
> say
> goodbye to privacy and security.

What's insecure about push?  And yeah, I guess if I didn't want Google
to know when some service wants to wake up my phone, I guess.  But I
don't see any value in that piece of data that is simply "wake up".

> I'll rather keep my phone charged.

Which is great, I guess if you are always sitting right beside a plug.

> I cannot use UDP, because I need TLS for security.

TLS does not require TCP.  https://tools.ietf.org/html/rfc6347

> One thing that I
> have noticed with UDP in Android is that when network connection is
> lost
> to my phone, my SIP proxy still thinks that the SIP UA is alive until
> its registration expires.

Yes.  I don't personally see this as a problem.

> With TCP or TLS, SIP proxy learns it
> immediately.

I guess that depends on what you mean by "network connection is lost". 
If you simply mean that the network dropped out from between the phone
and sip proxy, they no, the sip proxy wouldn't know.  It would simply
queue TCP packets for the remote as I described in my original message.

If some device on the path tells your sip proxy (i.e. through an ICMP
message) that the network has been lost, then yes it will know.  But
it's not actually normal or desirable for devices to return ICMP
messages for what should be a temporary outage.  It should rely on
TCP's reliability to allow TCP sessions to resume once the outage has
been restored.

Cheers,
b.

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