[mythtv] iPhone streaming implementation

[Chase Douglas]

On Mar 22, 2009, at 10:20 PM, Rob Smith wrote:
> On Sun, Mar 22, 2009 at 11:29 AM, Chase Douglas
> <chasedouglas.lists at gmail.com> wrote:
>> I am at a crossroad as I begin new work towards implementing an  
>> iPhone
>> frontend. The new iPhone OS 3.0 announced this past week allows for  
>> easier
>> streaming from MythTV. However, the implementation does not follow a
>> standard that exists today, so I am wondering how amenable the  
>> MythTV devs
>> are to incorporating code into MythTV just for the iPhone (though the
>> standard will become open once the iPhone 3.0 OS is released in  
>> June, and it
>> could be used by other devices if they so choose).
>
> There currently doesn't exist a streaming framework in myth at all. We
> do our 'streaming' of non-recorded media types by using ffmpeg or
> similar. For example, the current iphone streaming in mythweb requires
> a user job to pre-compress to mpeg4 and then it'll stream just fine.
>
> If you are looking for the backend to re-encode and stream, it's
> something I personally would love to have, and as for support, given
> we use libavcodec, add support upstream for the specific encoder type
> and it'll make it's way back to us. There's is a fair bit of work
> involved for both pieces, so it's not going to be something simple if
> you want it integrated in.

Actually, the ffmpeg pieces are currently there, although they seem to  
have been broken until a few weeks ago. I'm currently using ffmpeg 0.5  
and it works well. The issue has more to do with streaming support in  
MythTV given that the exact implementation may be iPhone specific.

I'm still trying to figure out what the best implementation is either  
way though. Given all the different ways to stream media even after  
it's transcoded (rtp, rtsp, mms, iPhone's method, each with some  
quirks depending on the specific intended client), maybe it would make  
the most sense to have a plugin architecture for streaming? Would that  
be frowned upon? To be clear, I think the transcoding could be  
implemented without too much issue (architecturally that is) without  
needing any plugins, but the streaming is where things may really get  
hairy if you try to force one-size-fits-all types of approaches.