Differences between DASH' AdaptationSets and the rx-player "Adaptation"

The RxPlayer defines an Adaptation object (also sometimes called Track) which follow as close as possible the concept of the AdaptationSet in the DASH protocol.

However, to answer practically to some of the features allowed by DASH while still respecting the DASH-IF "IOP", we had to take some (minor) freedom with our interpretation of it.

Merging of multiple AdaptationSets into a single Adaptation

The main difference is that all similar AdaptationSet which are marked as "seamlessly switchable" between one another are merged into a single Adaptation in the player.

Why do we do that

This "switchable" concept is for example used in cases were multiple encryption keys are present for different Representation (e.g. due to limitations coming from right holders).

The problem is that the DASH-IF tells us that all Representation in a given AdaptationSet have to use the same license. This means that in the aforementioned case, the concerned Representation have to be divided into multiple AdaptationSet. In a player, different AdaptationSet means different "tracks" and thus a player won't try to automatically switch between them.

This means that our adaptive algorithm won't be able to set the right quality and that the library user would have to manually manage that instead.

Fortunately, the DASH-IF IOP planned a work-around for that kind of situation: To allow a player to seamlessly switch between multiple AdaptationSets, the DASH-IF allows a specific node, called SupplementalProperty to be added as children of the concerned AdaptationSets (with a specific value).

However, this brings another set of issues in the rx-player, where this separation would lead to an excessively complicated API.

What do we do

We thus decided to "merge" the AdaptationSets into a single Adaptation if all those conditions are filled:

• they both support seamless-switching between one-another (i.e. they both contain a SupplementalProperty node with the right values)

• they represent the same type of content ("audio", "video" or "text")

• they are of the same language, if one (letter-for-letter in the manifest)

• they have the same accessibility information (e.g. both are closed captions or audio description for the visually impaired).

If any of these conditions is not filled, the concerned AdaptationSets stay separated and the player will not try to switch between them.