WebSocket interception

The last "known limitation": a ws://wss:// connection stops being
one-shot request/response the instant its 101 Switching Protocols
lands, and forward()'s normal write-response-then-record flow has no
way to represent that. Scoped to HTTP/1.1 client legs (HTTP/2 can't be
hijacked for raw post-response access the way HTTP/1.1 can, and
browsers open a dedicated HTTP/1.1 connection for WebSocket regardless
of the surrounding page's protocol, so this isn't a real-world gap).

internal/proxy/websocket.go decodes each RFC 6455 frame's opcode and
payload for capture while relaying the exact same raw bytes it read
unmodified - this is capture, not tampering, matching the rest of the
codebase's raw-bytes-as-source-of-truth stance. One row per frame, not
per reassembled message (fragmentation is rare in real-world
WebSocket traffic; not worth buffering an unbounded number of pending
fragments to handle it). forward() branches on a matching 101 into
handleWebSocketUpgrade, which hijacks the client connection, relays
the handshake response raw, records the upgrade request/response to
history normally, then relays frames bidirectionally into a new
ws_messages table - reachable from the TUI's detail view via `w`.

Found and fixed two real bugs by actually driving a WebSocket
connection through a running daemon, not by reading the code:
stripHopByHop was deleting Connection/Upgrade from every outgoing
request (correct for an ordinary request per RFC 7230, catastrophic
for one asking to upgrade - every WebSocket attempt silently became a
426); and the relay tore the whole connection down the instant either
side saw a close frame, before the peer's own close-frame reply could
be relayed back, producing an abrupt EOF instead of a clean close.

Verified live end to end on both paths a real client uses: ws://
(plain HTTP forward-proxying) against a Python websockets echo
server, and wss:// (CONNECT-tunneled, TLS-intercepted) against the
same server behind TLS - text, binary, and extended-length frames,
plus a full close handshake with both directions' close frames
present, confirmed via the actual bytes captured in ws_messages.

1 files changed, 21 insertions, 2 deletions

diff --git a/README.md b/README.md
index ad4db79..25fb28d 100644
--- a/README.md
+++ b/README.md
@@ -23,6 +23,9 @@ list of what's deliberately not implemented (and why), see
   independently on the client and upstream legs - a client that only
   speaks HTTP/1.1 and an origin that prefers HTTP/2 both work correctly
   in the same request.
+- **WebSocket**: `ws://`/`wss://` connections are relayed byte-for-byte
+  unmodified with every frame captured for display, not just the
+  upgrade handshake - see WebSocket below.
 - **History**: every request/response captured to SQLite. Raw wire
   bytes are preserved byte-for-byte on HTTP/1.1 legs (what request
   smuggling and parser-differential analysis actually needs); HTTP/2
@@ -258,7 +261,22 @@ response (display-only - never touches the stored or resent bytes), `c`
 mark/compare (same as the history list), `r`/`i` jump straight to
 Repeater/Intruder seeded from this entry, `e` exports the entry
 (request and response, raw bytes, plain text - type a path and press
-enter), `esc` back.
+enter), `w` views captured WebSocket messages if this entry's
+connection was upgraded (see WebSocket below), `esc` back.
+
+### WebSocket
+
+A `ws://` or `wss://` request that gets a matching `101 Switching
+Protocols` back stops being one-shot request/response - mitmux relays
+every frame byte-for-byte unmodified in both directions (this is
+capture, not tampering) while decoding each one's payload for display.
+Press `w` from an upgraded entry's detail view to see them: direction,
+opcode (text/binary/close/ping/pong), size, and a preview; `enter` on
+a row shows that frame's full decoded payload. One row per frame, not
+per reassembled logical message - a message fragmented across several
+frames (rare in real-world WebSocket traffic: JSON events, chat
+messages, game state are almost always single-frame) shows up as
+several rows rather than being stitched back together.
 
 ### Comparer
 
@@ -571,7 +589,8 @@ reasoning behind each:
   1000 requests per attack across all four modes
 - `mitmuxd -install-ca` prints per-OS trust-store install steps; it
   never runs them for you (see Quick start above for why)
-- No WebSocket interception
+- WebSocket messages are captured one row per frame, not reassembled
+  from fragments (rare in real-world traffic) - see WebSocket below
 - No active or passive vulnerability scanning, no plugin system - this
   is a manual-testing tool, not a scanner