Open haudiobe opened 5 years ago
Global streams would be helpful. SCTE 35 cue messages can be "open" for much longer than the segments are available (e.g. program boundaries), and may span several periods
I am unsure how the global events would work as you will lose synchronization to the media timeline in case periods are added, modified etc. During call Thomas raised similar suggestion introducing event stream continuity. Signalling continuity for events makes sense, might be the default even. What kind of events would be signalled in global event streams ?
Global events need to be timed off @availabilityStartTime, /methinks. The use case we have is "open" SCTE 35 events from previous periods where all segments are already out of availability window. Conceptually this can be viewed as "event history" of events with presentation time earlier than the current period. Additional use case can be when event duration spans multiple periods, b/c a period was created due to temporary loss of input transcoder input for a few 5 seconds.
On Wed, Sep 25, 2019 at 5:00 PM Rufael Mekuria notifications@github.com wrote:
I am unsure how the global events would work as you will lose synchronization to the media timeline in case periods are added, modified etc. During call Thomas raised similar suggestion introducing event stream continuity. Signalling continuity for events makes sense, might be the default even. What kind of events would be signalled in global event streams ?
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/Dash-Industry-Forum/AdInsertion/issues/45?email_source=notifications&email_token=AAGEYZM7RQYSG5GP6IIRMFTQLN4KXA5CNFSM4IKVDTSKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD7SHFTI#issuecomment-535065293, or mute the thread https://github.com/notifications/unsubscribe-auth/AAGEYZK7A6Z6KCRDRD63JPDQLN4KXANCNFSM4IKVDTSA .
The MPEG ad-hoc is picking up the global event stream work, should we defer this until that work is completed and pull in the result?
Sure!
On Fri, Jan 31, 2020, 00:08 Zachary Cava notifications@github.com wrote:
The MPEG ad-hoc is picking up the global event stream work, should we defer this until that work is completed and pull in the result?
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/Dash-Industry-Forum/AdInsertion/issues/45?email_source=notifications&email_token=AAGEYZLSKEEHFNBDBYVNBCDRAPE6RA5CNFSM4IKVDTSKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEKNW36Q#issuecomment-580611578, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAGEYZOKL2V73WNYQD2IQ2DRAPE6RANCNFSM4IKVDTSA .
How about just recommending use of presentationTimeOffset in event streams, and re-signalling events in each period ? adding presentationTimeOffset to EventStream was an important fix in amendment 5. Creating a new global event stream causes more confusion around this topic. I am also ok with having event stream continuity cross boundaries
To be constructive and try to progress this some proposed text for the document:
In case opportunity data crosses a next period, the following is recommended:
An EventStream Element with same schemeIdUri, value and timescale attributes is present in the next Period. The EventStream Element in the next period shares a common anchor, such as Unix Epoch 1-1-1970 or another shared anchor. This anchor is used to set the EventStream@presentationTimeOffset to align the EventStream to the start time of the period. Thus a period that start 10000 ticks on the timescale after the anchor would contain an EventStream@presentationTimeOffset of 10000. Events from previous periods that are still active in the current period are then signalled again using the same presentationTime, id and value as in one of the previous periods. In other words, they are repeated in this Period. In case the opportunity data is no longer active, such re-signalling is not needed.
DASH-IF IOP 2021/05/14:
@presentationTimeOffset
- similar as a media segment.
We need more study on Events across Event Boundaries. Two issues: