Add interface to disable DTLS SRTP

[CaptainNeil]

Summary

As the title describes, I would like to have an interface to disable DTLS SRTP, just like the enable_dtls_srtp defined in C++

Motivation

To analyze RTP payload.

Describe alternatives you've considered

I'm working on rtmp 2 webrtc, the workflow is as follow：

• rtmp receive origin aac data
• decode aac to pcm
• slice pcm to 10ms pcm list (the opus encoder requires)
• encode pcm to opus
• pack opus as an rtp package
• send rtp package to TrackLocalStaticRTP

Unfortunately, the received sound on the other end is noisy.
I saved the opus data before sending it to the RTPSender, and it sounded fine.

So I wanted to use Wireshark to see what happened, but the rtp payload was encrypted.

By searching at code，I could'n find a way to turn it off.

@Sean-Der

[HustCoderHu]

agreed, TLS should be configurable, coz encryption is not necessary under some circumstances, performance critical i.e.

[CaptainNeil]

The problem is similar as #2090

[zjzhang-cn]

agreed

[yyyar]

@CaptainNeil did you managed to solve the issue?
I have similar problem doing the same thing (aac->opus->webrtc) I've got choppy audio as described in #2090. I noticed 0 lost packets in webrtc-internals in chrome, but lots of concealedSamples and insertedSamplesForDeceleration.

[CaptainNeil]

@CaptainNeil did you managed to solve the issue? I have similar problem doing the same thing (aac->opus->webrtc) I've got choppy audio as described in #2090. I noticed 0 lost packets in webrtc-internals in chrome, but lots of concealedSamples and insertedSamplesForDeceleration.
@yyyar
I have solved this problem.
My problem is the noise caused by the rtp timestamp.
When I pack OPUS into an RTP packet, I did not increase the timestamp according to the audio size. Therefore, noise will be generated when webRTC rendering

[GodKingGitHub]

absolutely agreed

[Sean-Der]

I am in support of this!

What someone should do is add support for the NULL cipher to pion/srtp. Then via the SettingEngine both sides can request it.

As long as this requires the developer explicitly requesting it (via the SettingEngine) it would be great to have.