Create your own GitHub profile
Sign up for your own profile on GitHub, the best place to host code, manage projects, and build software alongside 50 million developers.
Sign up
-
Mozilla
- Australia
Pro
-
Mozilla
- Australia
Pinned
2,936 contributions in the last year
Activity overview
Contribution activity
May 2020
Created a pull request in quicwg/base-drafts that received 8 comments
A pacing algorithm
This isn't code, because I don't think we need to do any more than point at the leaky bucket algorithm, but it does provide more concrete advice th…
+33
−11
•
8
comments
- Revert "rename SERVER_BUSY to CONNECTION_REFUSED"
- DPLPMTU merge tweaks
- Citation rules for QUIC
- Better duplicate detection
- Forgeries likely exhaust TWO keys
- Add an example of a variable-length integer field
- Invariant independent
- ECN probing needs to imply that you stop
- Changelog for -28
- Update examples for -28
- Convert command-line arguments to integers
- Include max_udp_payload_size in packet size discussion
- s/alarm/timer/
- Restore missing period on rttvar definition
- One more semicolon
- Exception for Retire Prior To
- Use Section not section
- Quoting in recovery
- Use a list for TLS contributors too
- Define an anti-forgery limit
- Use impl AsRef rather than a generic parameter
- Probes don't cause more probes
- Use NSS option to remove EndOfEarlyData
- Make the commands default to the text-fixture database
- Preconnect for resumption
- Have the server send a session ticket
- Create an active method on Http3State
- Reduce strictness of assertions while closing
- Inspect 0-RTT state before attempting it
- Make sure to initialize the fixture
- Stop using wildcard imports
- Updated rust stable and now it formats macros
- Remove enable_loss_detection_timer boolean
- Sanitize application close frames
- Ignore negative ack delay
- Ignore kept lost packets
- Pacing
- Allow passing of String to close()
- Ignore timers while closing
- Allow ACK after PTO
- use TRANSPORT_PARAMETER_ERROR when authenticating connection IDs
- Integrate QUIC text from DPLPMTUD
- A suggested text about a push stream after a CANCEL_PUSH
- Servers should always send CANCEL_PUSH
- Split PTO calculation into a dedicated method
- rename SERVER_BUSY to CONNECTION_REFUSED
- Define an anti-forgery limit
- Add CID authentication to draft-28 changelog
- Scope of disable_active_migration
- Clarify text on pacing
- Clarify when the PTO may need to be recomputed and reset
- Include ack_delay when deciding with PN space to arm PTO for
- Move MUST not send bursts up in the pacing section
- Add OnPacketReceived to pseudocode
- Remove repeated sentence
- fix MUST not in CRYPTO frame handling
- Fix two typos in the transport draft
- Fix verb number and a typo in the QPACK draft
- Another pacing algorithm
- Exception for Retire Prior To
- Move a PTO section up
- QPACK acknowledgements
- [qpack] Table Flip
- A pacing algorithm
- Minor errors in new diagrams
- Some pull request reviews not shown.
- Add retry, resumption and zerortt tests
- Add test chacha20
- Drop packets with unknow versions
- Add new macro that logs in all circumstances
- Qpack encoder small refactor
- Use separate send_message/recv_message instead of transactions
- Proposed fix to #585
- Fix option conflict and change default values
- usize and Res simplifications in recv_stream.rs
- Reduce strictness of assertions while closing
- check_connection_events may actually change state of the connection.
- Sanitize application close frames
- Ignore kept lost packets
- The qpack encoder uses stream_send_atomic to send instructions
- Change stream_avail_send_space() to return usize instead of u64
- neqo-client: Handle UdpSocket::send() errors
- Pacing
- Add an encoder's max_table_size.
- Drop ack ranges
- Enforce maximum values for stream offsets and stream ID limits
- neqo-client: Update process_loop_old to call process_output() repeatedly
- Close a connection if critical streams are reset or a stop_sending is received
- Improve table.rs
Created an issue in quicwg/base-drafts that received 15 comments
Forgery limits on packet protection
tl;dr We need to recommend limits on the number of failed decryptions. Introduction At the QUIPS workshop one of the side discussions concentrated …
15
comments
- Remove EOED from TLS transcript too
- Send immediate ACK on PING
- Occasional STOP_SENDING from client
- Server API for 0-RTT
- Infinite PTO exchange
- Crash in closing state
- -d option to neqo-http3-server is conflicted
- Attempting 0-RTT when not possible
- Cloning RecoveryTokens
- Conditions where an ACK is not sent