Expand description
The same bulk-bytes treatment for a plain Vec<u8> field.
Used by cross-machine payloads that do not need the 128-byte Arrow
alignment. The in-tree users are the extension-envelope payloads —
InterDaemonEvent::ExtensionMessage, DaemonRequest::ExtensionRequest,
and DaemonReply::ExtensionReply — and, carried opaquely inside one of
those envelopes, the tensor-pool PeerMessage::Write { tensor_data }
field (dora_tensor_pool::protocol), whose receiver copies the bytes
into a mirror segment rather than decoding them zero-copy.
(InterDaemonEvent has no MemoryPoolWrite variant, and the tensor-pool
type is PeerMessage::Write, not MemoryPoolWrite.) The wire encoding is
identical to serde’s default Vec<u8> (postcard: varint length + raw
bytes), so this is a pure speedup with no version bump —
vec_encoding_is_unchanged_from_the_seq_form pins it.