[ppsp] Open issues around swift

[Arno Bakker <[e-mail]>]

Hi

the following is a list of the open issues around swift as raised on the 
PPSP mailing list from Nov 24-Dec 19. The list consists of my summary of 
the issues, as originally raised by the person noted between the square 
brackets, so any (mis)interpretations are mine, and I'll gladly correct 
them.

1. Need to clarify how swift deals with SVC/MDC. E.g. single swarm vs. 
multiple swarms. [Cruz,Zhang,Kiraly]

2. Do we need freedom regarding push/pull of chunk availability? [Wang]

3. Need to clarify that peer selection is not random, but policy 
controlled. [Yang]

4. Should we have explicit CHOKE/UNCHOKE messages? [Yang]

5. Do we need a policy that controls when chunk availability updates are 
pushed? [Yang]

6. Need to clarify how higher layers can control swift download process
for seeking, switch audio/subtitles, SVC. [Cruz]

7. Should we support pull-token or pull-push streaming, i.e.,
add more flexibility in the way chunks are retrieved? [Picconi]

8. Should we support View-Upload Decoupling? [Picconi]

9. Should we support substreams as a first class entity for SVC, 
multiple audio tracks? [Picconi]

10. Need to allow for multiple content-integrity protection schemes for 
live video. [Picconi]

11. Need to clarify that swift doesn't specify how content is
stored. [Cruz]

12. Should we support UNHINTing of chunk requests? [Picconi]

13. Allow for multiple content addressing schemes, explain dependencies 
of other swift features on bin addressing. [Kiraly,Zhang]

14. Need to clarify which features of swift depend on fixed-sized chunks 
and which are chunk-size independent. [Kiraly]

15. Need to add explanation to section 2 that this is an example 
behaviour, not requirement (regarding requesting disjunct sets of chunks 
from different peers). [Kiraly]

16. Do we need a policy that controls to *whom* chunk availability 
updates are pushed? [Kiraly]

17. How to formulate behaviour of Peer-Exchange messages? What is 
recent? [Kiraly]

18. Should we support transferring extra metadata about chunks and peers 
in the wire protocol? [Kiraly]

19. Should we support push and pull and OFFER-ACCEPT models of 
streaming, i.e., add more flexibility in the way chunks are retrieved? 
[Kiraly]

20. Evaluate security implications of PEX. [Bakker]

21. What information is required in on-the-wire protocol to allow for 
different congestion control mechanisms? [Stiemerling]

22. Should we define failure behaviour and semantics. E.g. When is a 
peer dead? [Stiemerling]

23. Should we define error codes? [Stiemerling]

24. Should we remove "RTP Profile for PPSP"? [Stiemerling]

25. Should we define how congestion control is signalled between peers? 
[Stiemerling]

26. Evaluate the handshake procedure if it needs strengthening against 
"state-building attacks"? [Stiemerling]

27. Need to clarify keep-alive mechanism. [Stiemerling]

28. Which transport should we support? UDP or TCP? [Stiemerling]

29. Do we need extra support for graceful and unexpected peer crashes? 
[Stiemerling]

30. Do we need a policy that controls *when* chunk availability updates 
are pushed? [Stiemerling]

31. Need to clarify why multiple HINTs in datagram. [Stiemerling]

32. Rename PEX_ADD to PEX_RES. [Stiemerling]

33. Need to clarify built-in NAT hole punching. [Stiemerling]

34. Need to clarify resource usage of Merkle hash trees (on-the-wire
overhead) [Bakker]

Did I miss anything?

Happy holidays!
       Arno