Things to do

[kazu-yamamoto]

The current target is draft-32.

Handshake

• [x] QUIC APIs in the Haskell TLS library
• [x] Hello retry request
• [x] New session ticket
• [x] Killing the handshake controller with timeout in the client side
• [x] Key udpate

Dispatching

• [x] NAT rebinding
• [x] Retry
• [x] Version negotiation
• [x] Path validation
• [x] Migration in the server side
• [x] Migration in the client side

Transport

• [x] Encoder and decoder for packets and frames
• [x] Basic APIs for connections
• [x] Resending packets with PSQ
• [x] Fragmentations and defragmentations
• [x] Dispatching based on CID
• [x] Closing a stream by sending the Fin bit vs closing the connection
• [ ] The spin bit
• [x] 0RTT
  • [ ] Server checks early data size
  • [ ] Client resends early data in 1-RTT if rejected
• [x] Amplification attack prevention (the three times rule)

HTTP/3 and QPACK

• [x] Toy client and server for HTTP/0.9 or ALPN=hq-draft_version
• [x] Toy client and server for HTTP/3 or ALPN=h3-draft_version
• [x] HTTP/3
• [x] QPACK
  • [x] Decoder
  • [x] Naive/Static encoder
  • [ ] Linear encoder (using dynamic table)
• [ ] Server push
• [x] Integrating to Warp (warp-quic)

Logging

• [x] Debug logging
• [x] qlog

Recovery

• [x] Flow control
  • [x] Max data
  • [x] Max stream data
  • [x] Max streams
• [x] Loss detection
• [x] Congestion control

[bjin]

Really appreciate your work on this project. While it's absolutely at an early stage to discuss this, I'd like to ask about the plan regarding the congestion control algorithm.

The draft recommends NewReno algorithm, but alternative algorithm like bbr is supported by some implementation including chromium and litespeed. So my question is, is basic congestion control planned to be implemented, and if so, will it be extensible to support more algorithm in the future?

[kazu-yamamoto]

@bjin Congestion control would be the very last part of my implementation. Very far away... Anyway, it should be extensible if users want!

[kazu-yamamoto]

@bjin I have implemented the recovery draft finally but the code is not stable yet.

[kazu-yamamoto]

I believe that the current code is quite stable.