[mythtv-users] Best way to integrate a Netflix feed into Myth ?

[Michael T. Dean]

On 08/22/2011 02:56 PM, Raymond Wagner wrote:
> On 8/22/2011 14:28, Brian J. Murrell wrote:
>> On 11-08-22 02:19 PM, Eric Sharkey wrote:
>>> It would be
>>> better if mythTV had a pass-through mode that disabled buffering to
>>> disk, eliminating the ability to pause/record/rewind but providing
>>> lower latency playback of the content and a way to pass through remote
>>> control commands from the frontend to the STB/appliance, but nothing
>>> like that exists right now.
>> I've previously wondered if that "write-to-disk-read-from-disk" latency
>> were actually necessary for something like live-tv.
>>
>> As I understand it, currently when you watch live-tv, myth starts
>> recording the tv channel and then starts playing back from the stream
>> it's writing to disk, incurring the latency.
> If everyone were actually streaming live, no, it wouldn't be necessary.
> The problem comes when you start trying to diverge from live.  You only
> want a couple seconds of video in memory at any given time.
>
> The best way to handle this would be to keep the two buffers, but allow
> them to be linked, one writing to the other.  If the read buffer on the
> backend diverges too far, it would decouple from the write buffer, and
> begin reading from the disk directly.  There would be other required
> changes.  You would need to force the ringbuffer on the frontend to read
> remotely from the backend during use of LiveTV.  You would need to
> rework the option to force streaming from the master backend, or only
> allow it to operate when not using LiveTV.  You would need to make sure
> the streaming and filesize updates were granular enough that the
> ringbuffer classes continue to function with such little readahead.
>
> The code itself is not a great deal of work, but actually making it
> function properly would be difficult.

And it's unlikely to have any useful effect on reducing latency, too.  
See the hundreds of threads where someone has suggested such an approach 
and the responses from the developers about where the latency actually lies.

Now, true passthrough (as from a framegrabber whose output is shown 
directly on screen) will reduce latency--but, as mentioned, can't be 
done in a networked situation.

Mike