Hi Everyone,
As we have discussed the idea of explicit Path IDs on the working group meeting 
in IETF 118, there're at least 20 people interested in the new design.
Marten and I have been working on a proposal about this idea for the working 
group. 
Please feel free to check the proposal and give comments on it, we look forward 
to hearing your suggestions.
Brief summary for the new proposal:
- (Most important) Use explicit Path IDs instead of CID sequence number to 
identify a path. Path IDs are used to address a path in control frames of 
Multi-path extension, e.g PATH_ABANDON, PATH_AVAILABLE and PATH_STANDBY.
- Path IDs are used as part of the AEAD encryption / decryption nonce instead 
of CID sequence number.
- Packet number space is per path, and it's bind to a Path Identifier. Packet 
number space doesn't change when CID rotation happens. 
- Transport parameter: Use "initial_max_paths" instead of "enable_multipath" to 
negotiate multi-path and max initial paths number. 
- New Frame Types:
 - Add MP_NEW_CONNECTION_ID frame which issues new CIDs associated with Path 
Identifiers. CID sequence and Retire Prior To are arranged per path. Add 
MP_RETIRE_NEW_CONNECTION_ID frame to retire a CID under a path.
 - Add MAX_PATHS frame to inform the peer about expanding the available Path 
IDs; PATH_ABANDON frame is used to abandon path and retire the associated Path 
ID (It's not a new frame but we add the retirement signal of Path ID in it).
Here is the proposal and original issue:
Proposal for explicit Path IDs (pull request): 
https://github.com/quicwg/multipath/pull/292 
<https://github.com/quicwg/multipath/pull/292 >
Complete version: 
https://github.com/quicwg/multipath/blob/dev/path_id/draft-ietf-quic-multipath.md
 
<https://github.com/quicwg/multipath/blob/dev/path_id/draft-ietf-quic-multipath.md
 >
Issue and previous discussion: https://github.com/quicwg/multipath/issues/214 
<https://github.com/quicwg/multipath/issues/214 >
Next steps: 
1. Involve more implementations to verify the design and do more interop tests. 
2. We are planning to have an interim meeting in January(2024) discussing about 
this proposal and whether the working group agree to accept the new design. 
If you are interested in, please contact us and participate in the interim 
meeting.
Quentin, Christian, Mirja, Yunfei, Kazuho, Furong and Michael have helped a lot 
on reviewing and improve this proposal.
Thanks for everyone and happy thanksgiving day!
Best Regards,
Yanmei Liu

Reply via email to