Porting the hlsdemux work in issue 698155 to 1.x

[Duncan Palmer]

I've been delving into this some more, and looking at a couple of
alternatives to porting the fluendo work to 1.x;
- Merging the changes which add support for i-frame playlists and
trick play. This looks quite do-able.
- Removing the internal bins, and hence support for alternative
renditions, from the fluendo element, and porting the result to 1.x.
This looks like quite a bad idea...

Support for i-frame playlists would require the byterange changes you
have in your branch Arnaud. Are you hoping to get these in? I see you
have a bug raised on your GstFragment improvements.

It looks like Fluendo's version of hlsdemux has a lot of fixes and
improvements related to streaming. I worry whether the 1.x element is
seeing a lot of use? there have been some nasty bugs there recently
which have taken some time to be noticed. But the addition of i-frame
playlists to 1.x would do me for now...



On 3 December 2013 14:03, Duncan Palmer <dpalmer at digisoft.tv> wrote:
> Hi Arnaud,
>
> Whilst I think your work will be quite useful for me, it doesn't
> include support for i-frame playlists, which is a feature I need.
> However, I see that Andoni's branch requires an I-frame playlist to be
> present to allow seeking, which seems a bit of a feature regression...
> It's interesting what you say about performance - will have to try it
> and see how I go. Andoni's branch also uses buffer lists, but not to
> hold the playlist. I'm not sure of the significance of that...
>
> Right now, I don't need support for alternative renditions, but I may
> down the track. I do wonder how much use these see in practice - do
> people typically use alternative renditions, or do they use
> alternative audio tracks and dvb subs in the transport stream instead?
>
> Dunk
>
> On 30 November 2013 16:33, Arnaud Vrac <rawoul at gmail.com> wrote:
>> On 29 nov. 2013, at 21:30, Duncan Palmer <dpalmer at digisoft.tv> wrote:
>>
>>> Thanks Arnaud. I'd be interested in looking at your proof of concept.
>>> I would prefer any work I do to be suitable for inclusion in the
>>> mainline.
>>
>> Hi Duncan,
>>
>> I had already ported the most important patches to master, I've posted them here:
>> https://github.com/rawoul/gst-plugins-bad/commits/hls
>>
>> This branch also includes some modifications to work with buffer lists for better performance. With all those patches the reference streams from Apple [1] work better.
>>
>> I've also posted my proof-of-concept demuxer here, which handle multiple renditions better:
>> https://github.com/rawoul/pochlsdemux
>>
>> A lot of features are missing though (see TODO), I might work on it again if the stream selection API is ever implemented.
>>
>> [1] https://developer.apple.com/streaming/#streaming
>>
>> --
>> Arnaud
>> _______________________________________________
>> gstreamer-devel mailing list
>> gstreamer-devel at lists.freedesktop.org
>> http://lists.freedesktop.org/mailman/listinfo/gstreamer-devel