38Pass
17Fail
3Unsupported
1Inconclusive
69%of 55 conclusive
Stack: ngtcp2 + nghttp3 ·
Language: C ·
Version: libcurl/8.23.0 ngtcp2/1.25.0
· Measured 2026-10-27
Badge for a README:
https://pqcrypta.com/conformance/badge/curl.svg
By class
Rolled up by what the clause requires rather than into one number. A client that ignores unknown extensions correctly but never emits the right error codes has a specific, nameable problem, and an average is exactly what hides it.
| Class | Passing means | Pass | Fail | Inconclusive |
|---|---|---|---|---|
| correctness | Rejected something invalid, with the code the RFC names. | 8 | 17 | 0 |
| discretionary | The RFC permits either behaviour; the report says which was chosen. | 11 | 0 | 1 |
| extensibility | Ignored something unknown and carried on. | 5 | 0 | 0 |
| interoperability | Correctly decoded something valid but demanding. | 6 | 0 | 0 |
| resilience | Recovered rather than giving up. | 8 | 0 | 0 |
Every result
| Test | Clause | Verdict | What happened |
|---|---|---|---|
| Version Negotiation offering a reserved version alongside v1 | RFC 9000 §6 | Pass | Responded correctly: sent Version Negotiation for 4 datagram(s); the client did not persist with an unsupported version. |
| Retry packet for source-address validation | RFC 9000 §8.1.2 | Pass | Responded correctly: echoed the Retry token and completed the handshake. |
| Reserved transport parameter (31·N+27) | RFC 9000 §18.1 | Pass | Ignored the unrecognised element and carried on: completed a handshake carrying a reserved transport parameter. |
| Unknown frame type in a 1-RTT packet | RFC 9000 §12.4 | Pass | Responded correctly: closed with FRAME_ENCODING_ERROR, which is the code §12.4 requires for a frame of unknown type. |
| NEW_CONNECTION_ID followed by RETIRE_CONNECTION_ID | RFC 9000 §5.1 | Pass | Recovered: rotated connection IDs and retired 5 of them. |
| Stateless reset | RFC 9000 §10.3 | Pass | Responded correctly: was sent a Stateless Reset and went quiet — nothing further arrived in the 3 seconds that followed, which is the draining period §10.3.1 requires (0 datagram(s) were already in flight when the reset went out and are not counted against it). |
| Deliberately tight MAX_DATA and MAX_STREAM_DATA | RFC 9000 §4 | Pass | Handled it: respected the window without sending a BLOCKED frame, which §4.1 permits. |
| ACK Frequency extension offered | draft-ietf-quic-ack-frequency | Pass | Handled it: ignored the extension, which the specification permits. |
| Packets marked ECT(0) | RFC 9000 §13.4 | Pass | Responded correctly: echoed ECN counts back in its ACKs: 7 ECT(0), 0 ECT(1), 0 ECN-CE. |
| Path MTU black hole above a threshold | RFC 9000 §14, RFC 8899 | Pass | Recovered: detected the black hole 1 time(s) and settled on a 1293-byte path MTU (0 datagram(s) swallowed). |
| Server-initiated PATH_CHALLENGE | RFC 9000 §8.2 | Pass | Responded correctly: answered with 1 PATH_RESPONSE. |
| 0-RTT rejected after the client sends early data | RFC 9001 §4.6.2 | Pass | Recovered and completed the follow-up request. |
| A second path offered mid-connection | draft-ietf-quic-multipath | Pass | Handled it: declined the multipath offer and stayed on one path, which is conformant. |
| Spontaneous 1-RTT key update | RFC 9001 §6.2 | Pass | Decoded it and completed the request. |
| Stream limits set to the minimum a request needs | RFC 9000 §4.6 | Pass | Tolerated it and continued. The specification permits this but does not require it; a client that rejected it would also be conformant. |
| One datagram in twelve dropped once the path is established | RFC 9000 §2.2, §13.3 | Pass | Recovered and completed the follow-up request. |
| Datagrams arriving from a second server address | RFC 9000 §9.6 | Pass | Tolerated it and continued. The specification permits this but does not require it; a client that rejected it would also be conformant. The client process had already exited 1002ms in, before this test's reaction point at 3015ms, so the silence after that is not evidence either way. |
| Transport parameter carrying a value the specification forbids | RFC 9000 §7.4, §18.2 | Fail | abandoned the handshake without a CONNECTION_CLOSE. It saw the parameter -- that travels in the handshake -- and §7.4 requires treating it as a connection error of type TRANSPORT_PARAMETER_ERROR, which RFC 9000 §10.2 signals in a CONNECTION_CLOSE. None was sent: this client runs on the endpoint's own host, where nothing is lost in transit. |
| 425 (Too Early) in answer to a request sent as early data | RFC 8470 §5.2 | Pass | Tolerated it and continued. The specification permits this but does not require it; a client that rejected it would also be conformant. |
| Path marking packets CE, not just ECT(0) | RFC 9000 §13.4 | Pass | Responded correctly: reported the congestion back: 7 ECN-CE among 0 ECT(0) and 0 ECT(1), after 7 datagram(s) were marked. |
| A second key update after the first is acknowledged | RFC 9001 §6.1, §6.5 | Pass | Decoded it and completed the request. |
| No stream credit at first, then MAX_STREAMS mid-connection | RFC 9000 §4.6 | Unsupported | does not wait for stream credit. This port grants no bidirectional stream until 700ms in; the client closed the connection before then without opening one, so it never exceeded the limit -- which is all §4.6 requires -- and never made its request. |
| Reserved SETTINGS identifier (0x1f·N+0x21) | RFC 9114 §7.2.4.1 | Pass | Ignored the unrecognised element and completed the follow-up request. The client process had already exited 1102ms in, before this test's reaction point at never observed, so the silence after that is not evidence either way. |
| Reserved frame type on the response stream | RFC 9114 §7.2.8 | Pass | Ignored the unrecognised element and completed the follow-up request. |
| Unidirectional stream with a reserved stream type | RFC 9114 §6.2.3 | Pass | Ignored the unrecognised element and completed the follow-up request. The client process had already exited 1102ms in, before this test's reaction point at never observed, so the silence after that is not evidence either way. |
| SETTINGS containing the same identifier twice | RFC 9114 §7.2.4 | Pass | Tolerated it and continued. The specification permits this but does not require it; a client that rejected it would also be conformant. The client process had already exited 1102ms in, before this test's reaction point at never observed, so the silence after that is not evidence either way. |
| A DATA frame on the control stream | RFC 9114 §7.2.1 | Fail | Objected at the QUIC layer (rejected at the QUIC layer with INTERNAL_ERROR) instead of signalling the HTTP/3 error. The violation was detected and rejected, which is the right instinct; §8.1 requires an HTTP/3 connection error to be carried in an application close, and this test names 0x105. A transport close carries no HTTP/3 code at all, so a peer has no way to learn what was wrong. |
| Control stream whose first frame is not SETTINGS | RFC 9114 §6.2.1 | Fail | Objected at the QUIC layer (rejected at the QUIC layer with INTERNAL_ERROR) instead of signalling the HTTP/3 error. The violation was detected and rejected, which is the right instinct; §8.1 requires an HTTP/3 connection error to be carried in an application close, and this test names 0x10a. A transport close carries no HTTP/3 code at all, so a peer has no way to learn what was wrong. |
| A second control stream opened by the server | RFC 9114 §6.2.1 | Fail | Objected at the QUIC layer (rejected at the QUIC layer with INTERNAL_ERROR) instead of signalling the HTTP/3 error. The violation was detected and rejected, which is the right instinct; §8.1 requires an HTTP/3 connection error to be carried in an application close, and this test names 0x103. A transport close carries no HTTP/3 code at all, so a peer has no way to learn what was wrong. |
| Field lines referencing dynamic table insertions | RFC 9204 §4.3.3, §4.5.2 | Unsupported | the client advertised SETTINGS_QPACK_MAX_TABLE_CAPACITY of 0 and SETTINGS_QPACK_BLOCKED_STREAMS of 0, which forbid the server's encoder from using the dynamic table at all. |
| Huffman-coded field lines with maximal padding | RFC 9204 §4.1.2, RFC 7541 §5.2 | Pass | Decoded it and completed the request. |
| Field section larger than the client's advertised maximum | RFC 9114 §4.2.2 | Pass | Decoded it and completed the request. |
| Trailing field section after the body | RFC 9114 §4.1 | Pass | Decoded it and completed the request. |
| 103 Early Hints before the final response | RFC 9110 §15.2, RFC 8297 | Pass | Recovered and completed the follow-up request. |
| GOAWAY sent mid-connection | RFC 9114 §5.2 | Pass | Recovered and completed the follow-up request. The client process had already exited 1104ms in, before this test's reaction point at never observed, so the silence after that is not evidence either way. |
| MAX_PUSH_ID sent by the server | RFC 9114 §7.2.7 | Fail | Objected at the QUIC layer (rejected at the QUIC layer with INTERNAL_ERROR) instead of signalling the HTTP/3 error. The violation was detected and rejected, which is the right instinct; §8.1 requires an HTTP/3 connection error to be carried in an application close, and this test names 0x105. A transport close carries no HTTP/3 code at all, so a peer has no way to learn what was wrong. |
| SETTINGS frame on a request stream | RFC 9114 §7.2.4 | Fail | Objected at the QUIC layer (rejected at the QUIC layer with INTERNAL_ERROR) instead of signalling the HTTP/3 error. The violation was detected and rejected, which is the right instinct; §8.1 requires an HTTP/3 connection error to be carried in an application close, and this test names 0x105. A transport close carries no HTTP/3 code at all, so a peer has no way to learn what was wrong. |
| DATA frame before any HEADERS on the response stream | RFC 9114 §4.1 | Fail | Objected at the QUIC layer (rejected at the QUIC layer with INTERNAL_ERROR) instead of signalling the HTTP/3 error. The violation was detected and rejected, which is the right instinct; §8.1 requires an HTTP/3 connection error to be carried in an application close, and this test names 0x105. A transport close carries no HTTP/3 code at all, so a peer has no way to learn what was wrong. |
| CANCEL_PUSH for a push ID that was never promised | RFC 9114 §7.2.3 | Fail | Objected at the QUIC layer (rejected at the QUIC layer with INTERNAL_ERROR) instead of signalling the HTTP/3 error. The violation was detected and rejected, which is the right instinct; §8.1 requires an HTTP/3 connection error to be carried in an application close, and this test names 0x108. A transport close carries no HTTP/3 code at all, so a peer has no way to learn what was wrong. |
| Field section that blocks until the encoder stream catches up | RFC 9204 §2.1.2, §2.2.1 | Unsupported | the client advertised SETTINGS_QPACK_MAX_TABLE_CAPACITY of 0 and SETTINGS_QPACK_BLOCKED_STREAMS of 0, which forbid the server's encoder from using the dynamic table at all. |
| Response stream reset mid-body with H3_REQUEST_CANCELLED | RFC 9114 §8, §4.1 | Pass | Tolerated it and continued. The specification permits this but does not require it; a client that rejected it would also be conformant. |
| PRIORITY_UPDATE sent by the server | RFC 9218 §7.2 | Pass | Rejected it (rejected at the QUIC layer with INTERNAL_ERROR). The specification permits this but does not require it; a client that ignored it would also be conformant. |
| SETTINGS_ENABLE_CONNECT_PROTOCOL with a value outside 0 and 1 | RFC 9220 §3, RFC 8441 §3 | Pass | Rejected it (rejected at the QUIC layer with INTERNAL_ERROR). The specification permits this but does not require it; a client that ignored it would also be conformant. |
| PUSH_PROMISE for a push the client never allowed | RFC 9114 §7.2.5, §4.6 | Fail | Objected at the QUIC layer (rejected at the QUIC layer with INTERNAL_ERROR) instead of signalling the HTTP/3 error. The violation was detected and rejected, which is the right instinct; §8.1 requires an HTTP/3 connection error to be carried in an application close, and this test names 0x108. A transport close carries no HTTP/3 code at all, so a peer has no way to learn what was wrong. |
| SETTINGS_H3_DATAGRAM with a value that is neither 0 nor 1 | RFC 9297 §2.1.1 | Fail | Objected at the QUIC layer (rejected at the QUIC layer with INTERNAL_ERROR) instead of signalling the HTTP/3 error. The violation was detected and rejected, which is the right instinct; §8.1 requires an HTTP/3 connection error to be carried in an application close, and this test names 0x109. A transport close carries no HTTP/3 code at all, so a peer has no way to learn what was wrong. |
| Encoder stream setting a dynamic table capacity above the client's limit | RFC 9204 §4.3.1, §6 | Fail | Objected at the QUIC layer (rejected at the QUIC layer with INTERNAL_ERROR) instead of signalling the HTTP/3 error. The violation was detected and rejected, which is the right instinct; §8.1 requires an HTTP/3 connection error to be carried in an application close, and this test names 0x201. A transport close carries no HTTP/3 code at all, so a peer has no way to learn what was wrong. |
| A second GOAWAY naming a larger identifier than the first | RFC 9114 §5.2 | Fail | Objected at the QUIC layer (rejected at the QUIC layer with INTERNAL_ERROR) instead of signalling the HTTP/3 error. The violation was detected and rejected, which is the right instinct; §8.1 requires an HTTP/3 connection error to be carried in an application close, and this test names 0x108. A transport close carries no HTTP/3 code at all, so a peer has no way to learn what was wrong. |
| Push stream opened for a push nobody allowed | RFC 9114 §6.2.2 | Fail | Objected at the QUIC layer (rejected at the QUIC layer with INTERNAL_ERROR) instead of signalling the HTTP/3 error. The violation was detected and rejected, which is the right instinct; §8.1 requires an HTTP/3 connection error to be carried in an application close, and this test names 0x108. A transport close carries no HTTP/3 code at all, so a peer has no way to learn what was wrong. |
| Field line indexing a static table entry that does not exist | RFC 9204 §3.1, §4.5.2 | Fail | Objected at the QUIC layer (rejected at the QUIC layer with INTERNAL_ERROR) instead of signalling the HTTP/3 error. The violation was detected and rejected, which is the right instinct; §8.1 requires an HTTP/3 connection error to be carried in an application close, and this test names 0x200. A transport close carries no HTTP/3 code at all, so a peer has no way to learn what was wrong. |
| Datagrams delivered out of order | RFC 9000 §2.2 | Pass | Recovered and completed the follow-up request. |
| Encoder instruction naming a static index that does not exist | RFC 9204 §3.1, §4.3.2 | Fail | Objected at the QUIC layer (rejected at the QUIC layer with INTERNAL_ERROR) instead of signalling the HTTP/3 error. The violation was detected and rejected, which is the right instinct; §8.1 requires an HTTP/3 connection error to be carried in an application close, and this test names 0x201. A transport close carries no HTTP/3 code at all, so a peer has no way to learn what was wrong. |
| Server negotiates only the post-quantum hybrid X25519MLKEM768 | RFC 8446 §4.1.4, draft-ietf-tls-hybrid-design | Pass | Decoded it and completed the request. |
| Server negotiates only classical X25519 against a hybrid offer | RFC 8446 §4.1.1, draft-ietf-tls-hybrid-design §5 | Pass | Tolerated it and continued. The specification permits this but does not require it; a client that rejected it would also be conformant. |
| Hybrid key share large enough to split the Initial across packets | RFC 9000 §8.1, §14.1, RFC 9001 §4.4 | Pass | Recovered and completed the follow-up request. |
| Server selects a key exchange group the client never offered | RFC 8446 §4.1.3, §4.2.8, RFC 9001 §4.8 | Fail | abandoned the handshake without any alert. The group was not accepted, but §4.1.3 requires the abort to carry illegal_parameter, and RFC 9001 §4.8 delivers an alert -- a generic one is permitted -- in a CONNECTION_CLOSE. None was sent: this client runs on the endpoint's own host, where nothing is lost in transit. |
| Hybrid key share with an intact X25519 half and a corrupt ML-KEM half | draft-ietf-tls-hybrid-design §3.2, RFC 8446 §4.1.3 | Pass | Responded correctly: did not complete the handshake (timed out). The shared secret is both halves through the key schedule, so a corrupt ML-KEM half must break it -- and this client did not fall back to the intact X25519 half. |
| Hybrid key share whose length does not match the group named with it | draft-kwiatkowski-tls-ecdhe-mlkem §3.1.2, RFC 8446 §6.2, RFC 9001 §4.8 | Fail | abandoned the handshake without any alert. §3.1.2 requires the abort to carry illegal_parameter, and RFC 9001 §4.8 delivers an alert -- a generic one is permitted -- in a CONNECTION_CLOSE. None was sent: this client runs on the endpoint's own host, where nothing is lost in transit. |
| Post-quantum certificate chain, compressed per RFC 8879 | RFC 8879 §4, RFC 8446 §4.4.2 | Inconclusive | The handshake ended (aborted by peer: the endpoint encountered an internal error and cannot continue with the connection) without a CONNECTION_CLOSE carrying a TLS alert, so nothing shows how far into the chain the client got. The client process had already exited 100ms in, before this test's reaction point at never observed, so the silence after that is not evidence either way. |
| GREASE named group a client must tolerate | RFC 8701 §4, RFC 8446 §4.2.7 | Pass | Ignored the unrecognised element and completed the follow-up request. |