testing ISOBMFF level features that are independent of codec

[jpiesing]

There are a number of ISOBMFF level features that are independent of codec and IMHO should be tested. There are a number of possibilities for how this could be organised;

1. Included in the standard test streams for each codec
2. Included in the standard test streams for one codec
3. Dedicated test streams

Examples of these features/tests include;
a. emsg box (v0 and v1)
b. negative composition time offset.
c. that undefined ISOBMFF boxes are ignored as expected/required & don't break things
d. edit lists

[jpiesing]

Now the mezzanine content could be built any time, it's important to assemble a list of the test content that's needed. We have a list of the AVC codec options. Other codecs will have other options. We need to consider the list of codec-independent options.

[haudiobe]

Agree with this. We should add a section on media independent test streams and collect the relevant data there. At the same time we should create this content with the simplest content that is expected to play everywhere, i.e. AVCHD and AACCore. Remove the specific issues from the other media profiles unless specifically justified.

[dsilhavy]

As far as I understand there are also different ways to map the segments on the MSE buffer timeline. Does this test include:

• Segments which define sidx.ept (EPT) and tfdt.baseMediaDecodetTime (BMDT) + composition offset.
• Would it be interesting to also check what happens if EPT and BMDT are not compliant? Which one takes precedence?
• Agree with Jon that c) Ignore undefined boxes is a very important test

[jpiesing]

As far as I understand there are also different ways to map the segments on the MSE buffer timeline. Does this test include:

* Segments which define sidx.ept (EPT) and tfdt.baseMediaDecodetTime (BMDT) + composition offset.

* Would it be interesting to also check what happens if EPT and BMDT are not compliant? Which one takes precedence?

@dsilhavy How many distinct tests are you suggesting?

[jpiesing]

For edit lists, there is a long discussion in MPEG about their use with audio which might help inform what should be tested. MPEGGroup/CMAF#24

[gitwjr]

Add a clause on how valid is the coverage, in particular with audio with respect to edit lists.
Defer to future implementation.