multi-frame transport

[bmegli]

Related to hardware-video-streaming#5

The idea is to make MLSP aware of transporting multiple logical subframes in a single frame.

The gains are:

• avoids unnecessary copies when preparing packets
• avoids unnecessary complexity on library user side

The disadvantages are:

• adds complexity in MLSP implementation
• adds complexity in MLSP interface
• needs protocol change (information about frame structure)

[bmegli]

needs protocol change (information about frame structure)

The frame structure may be encoded only with first (or last) packet and needs:

• subframes number (or sizes followed by 0)
• subframes sizes
• optionally later subframe semantics (what it caries)

[bmegli]

The frame structure may be encoded in the current payload:

• the current packet level header will be unchanged
• the paylod will begin with frame structure description
• the things that change are:
  • preparing packets (encode structure at the beginning of payload)
  • returning frames to user (decode structure after collecting frame)

Now, there is no need to copy any data to subframes when decoding.
It is enough to move the pointers to relevant sections.