refactor!: one file_chunk and index_sync encoder for every transport

`file_chunk` and `index_sync` were each built twice, once per transport, and
the two copies did not agree. WebRTC sent binary, unsigned chunks carrying a
`file_id`; QUIC sent base64 fields, two BLAKE3 hashes, a per-chunk Ed25519
signature and no `file_id`. `index_sync` was plain entries on one transport
and a `GroupIndex.serialize()` envelope on the other. One message type, two
shapes, one consumer each, and nothing that failed when they drifted — finding
C6 one size down, in the two places the handshake unification did not reach.

Phase 9.15 moved WebRTC to the binary format and dropped the per-chunk
signature; the QUIC encoder was never brought along. It is dropped here rather
than reintroduced: the AES-GCM tag authenticates the ciphertext under a
GEK-derived key, and since C3 the node authenticates itself once in the
handshake instead of once per megabyte.

`meshbay_common.protocol` now owns the chunk codec (`chunk_ciphertext`,
`file_chunk_wire`, `file_chunk_plaintext`) and `meshbay_node/transport/wire.py`
the index builder, which also absorbs the delta the daemon used to hand-build.
`test_transport_wire_parity.py` fails if either server grows its own copy back.

`ChunkRequest`/`ChunkResponse` are deleted. `ChunkResponse` described the QUIC
half while reading like the contract for both, which is what made the fork hard
to see at all.

BREAKING CHANGE: MNP 0.15 changes the encoding of `file_chunk` and `index_sync`
on the QUIC transport. The WebRTC shapes are byte for byte unchanged and no
QUIC client ships, which is why this is a MINOR bump; a deployed QUIC peer
would have made it MAJOR.

Also fixes a test fixture that put a `Path` where the daemon puts a `RootSet`.
Nothing caught it: the old QUIC index handler never touched `roots`, and
`entry_abs_path` fell through `Path.resolve(strict=...)`, reading the virtual
path as a truthy flag and returning the right file by accident.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01AsoWC3GmhNdwVFomW3QjH3

1 files changed, 84 insertions, 16 deletions

diff --git a/packages/meshbay-common/src/meshbay_common/protocol.py b/packages/meshbay-common/src/meshbay_common/protocol.py
index c789287..4d2dd9d 100644
--- a/packages/meshbay-common/src/meshbay_common/protocol.py
+++ b/packages/meshbay-common/src/meshbay_common/protocol.py
@@ -15,6 +15,11 @@ from typing import Any
 # second copy here said "0.1" while every message on the wire carried "0.2".
 # Nothing imported it, which is the only reason it was harmless.
 from meshbay_common import MNP_VERSION, MHP_VERSION   # noqa: F401  (re-export)
+from meshbay_common.webcrypto import (
+    chunk_key_aes,
+    decrypt_chunk_aes,
+    encrypt_chunk_aes,
+)
 
 
 # ── MNP message types ─────────────────────────────────────────────────────────
@@ -236,21 +241,84 @@ class IndexDelta:
     updates:      list[IndexEntry] = field(default_factory=list)
 
 
-# ── Chunk request/response ────────────────────────────────────────────────────
+# ── File chunk ────────────────────────────────────────────────────────────────
+#
+# One encoder and one decoder, for every transport.
+#
+# There used to be two. WebRTC moved to a binary wire format in Phase 9.15 (base64
+# costs 33%) and dropped the per-chunk Ed25519 signature with it; the QUIC encoder
+# was not brought along, so `file_chunk` meant two different messages depending on
+# which transport carried it — base64 fields, hashes and a signature on one, raw
+# bytes and a `file_id` on the other. Nothing in the type name said which, and the
+# dataclass that used to sit here described only the QUIC half while reading like
+# the contract for both. That is finding C6 one size down: two implementations of
+# one message, free to drift, with a test for neither.
+#
+# Why no per-chunk signature: the AES-GCM tag already authenticates the ciphertext
+# under a key derived from the GEK, which only group members hold, and since C3 the
+# node authenticates itself in the handshake and is pinned by the client. A
+# signature per chunk re-proved, once per megabyte, what the session established
+# once. (`sign_chunk` still exists in `crypto.py` — it signs the serialized index
+# envelope, which is a different artifact; see `indexer/group_index.py`.)
 
-@dataclass
-class ChunkRequest:
-    file_id:     str   # blake3 hash of file (hex)
-    chunk_index: int
 
-@dataclass
-class ChunkResponse:
-    chunk_index:    int
-    plaintext_size: int
-    nonce_b64:      str
-    ct_b64:         str
-    ct_hash_b64:    str
-    pt_hash_b64:    str
-    sig_b64:        str
-    pk_node_b64:    str
-    file_hash_b64:  str
+def chunk_ciphertext(
+    gek: bytes,
+    plaintext: bytes,
+    chunk_index: int,
+    file_hash: bytes,
+) -> tuple[bytes, bytes]:
+    """
+    `(nonce, ciphertext)` for one chunk.
+
+    Split out because `stream_data` encrypts exactly like `file_chunk` — same key
+    derivation, indexed by segment instead of by chunk — and differs only in the
+    message it lands in. It used to do so through its own copy of these two lines.
+    """
+    ckey = chunk_key_aes(gek, file_hash, chunk_index)
+    return encrypt_chunk_aes(ckey, plaintext)
+
+
+def file_chunk_wire(
+    gek: bytes,
+    plaintext: bytes,
+    chunk_index: int,
+    file_hash: bytes,
+    file_id: str = "",
+) -> dict:
+    """
+    Encrypt one chunk and build the `file_chunk` message.
+
+    `file_hash` is the raw content hash the chunk key is derived from; `file_id` is
+    the same value hex-encoded, echoed so a client running several downloads at once
+    can tell whose reply arrived. Serving a thumbnail or a cached transcode passes
+    the cache blob's own hash for both.
+    """
+    nonce, ct = chunk_ciphertext(gek, plaintext, chunk_index, file_hash)
+    return {
+        "type": MNP.FILE_CHUNK,
+        "v": MNP_VERSION,
+        "file_id": file_id,
+        "chunk_index": chunk_index,
+        "plaintext_size": len(plaintext),
+        "nonce": nonce,
+        "ct": ct,
+    }
+
+
+def file_chunk_plaintext(
+    gek: bytes,
+    msg: dict,
+    file_hash: bytes | None = None,
+) -> bytes:
+    """
+    Decrypt a `file_chunk`. Raises `InvalidTag` if the ciphertext was tampered with.
+
+    `file_hash` defaults to the message's own `file_id`, which is what a client that
+    asked for a whole file already has. Pass it explicitly only when the caller knows
+    better than the peer does.
+    """
+    if file_hash is None:
+        file_hash = bytes.fromhex(msg["file_id"])
+    ckey = chunk_key_aes(gek, file_hash, msg["chunk_index"])
+    return decrypt_chunk_aes(ckey, msg["nonce"], msg["ct"])