[pulseaudio-discuss] Suggested improvements for the raop module

[Bryan Gleeson]

Colin,

>Awesome, thanks for your analysis. I'm sure this will help smooth things
>out. As it's your work, would you be able to create a git-formatted
>patch such that you get the proper credit?

Sure - I'll need to familiarize myself with the details but it should be no problem. 

On the sending buffer size issue it seems common practice, and also the easiest approach, is not to set any sndbuf size in the socket application itself. The OS default size (the tcp_wmem sysctl parameter) will be picked up as a result; also the OS may automatically adapt this under load conditions and specifying a fixed value disables this automatic tuning. So providing the default OS value is large enough, which in general I think to be the case, simply removing the call to set the buffer size should be adequate, but I'll look into this a bit more.

On the receive buffer size (via SDP fmtp param) I'll also do a bit more testing to see what the behaviour is with different values.

> One last point - are the specs for raop v1 and raop v2 published
> anywhere or have these protocols been reverse engineered?

>>We've been trying to reverse engineer it...

Yes I now understand that this is to Apple just like another version of their iphone dock connector spec - a separate (from the App SDK) licensing arrangement with a number of CE hardware vendors to allow development of compatible speakers, hi-fi components, etc. Already Pioneer, Denon, JBL etc have AirPlay compatible devices. As such it is very unlikely that where will ever be an open publicly available spec since Apple will want to keep tight control on how AirPlay is used in any commercial products, particularly on the video streaming side. 

Regards,
Bryan