Files
syncthing/proto/bep/bep.proto
Jakob Borg b1c8f88a44 chore: remove weak hashing which does not pull its weight (#10005)
We've had weak/rolling hashing in the code for quite a while. It was a
popular request for a while, based on the belief that rsync does this
and we should too. However, the benefit is quite small; we save on
average about 0.8% of transferred blocks over the population as a whole:

<img width="974" alt="Screenshot 2025-03-28 at 17 09 02"
src="https://github.com/user-attachments/assets/bbe10dea-f85e-4043-9823-7cef1220b4a2"
/>

This would be fine if the cost was comparably low, however the downside
of attempting rolling hash matching is that we (by default) do a
complete file read on the destination in order to look for matches
before we starting pulling blocks for the file. For any larger file this
means a sometimes long, I/O-intensive pause before the file starts
syncing, for usually no benefit.

I propose we simply rip off the bandaid and save the effort.
2025-03-29 13:21:10 +01:00

254 lines
5.2 KiB
Protocol Buffer

syntax = "proto3";
package bep;
// --- Pre-auth ---
message Hello {
string device_name = 1;
string client_name = 2;
string client_version = 3;
int32 num_connections = 4;
int64 timestamp = 5;
}
// --- Header ---
message Header {
MessageType type = 1;
MessageCompression compression = 2;
}
enum MessageType {
MESSAGE_TYPE_CLUSTER_CONFIG = 0;
MESSAGE_TYPE_INDEX = 1;
MESSAGE_TYPE_INDEX_UPDATE = 2;
MESSAGE_TYPE_REQUEST = 3;
MESSAGE_TYPE_RESPONSE = 4;
MESSAGE_TYPE_DOWNLOAD_PROGRESS = 5;
MESSAGE_TYPE_PING = 6;
MESSAGE_TYPE_CLOSE = 7;
}
enum MessageCompression {
MESSAGE_COMPRESSION_NONE = 0;
MESSAGE_COMPRESSION_LZ4 = 1;
}
// --- Actual messages ---
// Cluster Config
message ClusterConfig {
repeated Folder folders = 1;
bool secondary = 2;
}
message Folder {
string id = 1;
string label = 2;
bool read_only = 3;
bool ignore_permissions = 4;
bool ignore_delete = 5;
bool disable_temp_indexes = 6;
bool paused = 7;
repeated Device devices = 16;
}
message Device {
bytes id = 1;
string name = 2;
repeated string addresses = 3;
Compression compression = 4;
string cert_name = 5;
int64 max_sequence = 6;
bool introducer = 7;
uint64 index_id = 8;
bool skip_introduction_removals = 9;
bytes encryption_password_token = 10;
}
enum Compression {
COMPRESSION_METADATA = 0;
COMPRESSION_NEVER = 1;
COMPRESSION_ALWAYS = 2;
}
// Index and Index Update
message Index {
string folder = 1;
repeated FileInfo files = 2;
int64 last_sequence = 3; // the highest sequence in this batch
}
message IndexUpdate {
string folder = 1;
repeated FileInfo files = 2;
int64 last_sequence = 3; // the highest sequence in this batch
int64 prev_sequence = 4; // the highest sequence in the previous batch
}
message FileInfo {
// The field ordering here optimizes for struct size / alignment --
// large types come before smaller ones.
string name = 1;
int64 size = 3;
int64 modified_s = 5;
uint64 modified_by = 12;
Vector version = 9;
int64 sequence = 10;
repeated BlockInfo blocks = 16;
bytes symlink_target = 17;
bytes blocks_hash = 18;
bytes encrypted = 19;
FileInfoType type = 2;
uint32 permissions = 4;
int32 modified_ns = 11;
int32 block_size = 13;
PlatformData platform = 14;
// The local_flags fields stores flags that are relevant to the local
// host only. It is not part of the protocol, doesn't get sent or
// received (we make sure to zero it), nonetheless we need it on our
// struct and to be able to serialize it to/from the database.
uint32 local_flags = 1000;
// The version_hash is an implementation detail and not part of the wire
// format.
bytes version_hash = 1001;
// The time when the inode was last changed (i.e., permissions, xattrs
// etc changed). This is host-local, not sent over the wire.
int64 inode_change_ns = 1002;
// The size of the data appended to the encrypted file on disk. This is
// host-local, not sent over the wire.
int32 encryption_trailer_size = 1003;
bool deleted = 6;
bool invalid = 7;
bool no_permissions = 8;
}
enum FileInfoType {
FILE_INFO_TYPE_FILE = 0;
FILE_INFO_TYPE_DIRECTORY = 1;
FILE_INFO_TYPE_SYMLINK_FILE = 2 [deprecated = true];
FILE_INFO_TYPE_SYMLINK_DIRECTORY = 3 [deprecated = true];
FILE_INFO_TYPE_SYMLINK = 4;
}
message BlockInfo {
bytes hash = 3;
int64 offset = 1;
int32 size = 2;
reserved 4;
}
message Vector {
repeated Counter counters = 1;
}
message Counter {
uint64 id = 1;
uint64 value = 2;
}
message PlatformData {
UnixData unix = 1;
WindowsData windows = 2;
XattrData linux = 3;
XattrData darwin = 4;
XattrData freebsd = 5;
XattrData netbsd = 6;
}
message UnixData {
// The owner name and group name are set when known (i.e., could be
// resolved on the source device), while the UID and GID are always set
// as they come directly from the stat() call.
string owner_name = 1;
string group_name = 2;
int32 uid = 3;
int32 gid = 4;
}
message WindowsData {
// Windows file objects have a single owner, which may be a user or a
// group. We keep the name of that account, and a flag to indicate what
// type it is.
string owner_name = 1;
bool owner_is_group = 2;
}
message XattrData {
repeated Xattr xattrs = 1;
}
message Xattr {
string name = 1;
bytes value = 2;
}
// Request
message Request {
int32 id = 1;
string folder = 2;
string name = 3;
int64 offset = 4;
int32 size = 5;
bytes hash = 6;
bool from_temporary = 7;
int32 block_no = 9;
reserved 8;
}
// Response
message Response {
int32 id = 1;
bytes data = 2;
ErrorCode code = 3;
}
enum ErrorCode {
ERROR_CODE_NO_ERROR = 0;
ERROR_CODE_GENERIC = 1;
ERROR_CODE_NO_SUCH_FILE = 2;
ERROR_CODE_INVALID_FILE = 3;
}
// DownloadProgress
message DownloadProgress {
string folder = 1;
repeated FileDownloadProgressUpdate updates = 2;
}
message FileDownloadProgressUpdate {
FileDownloadProgressUpdateType update_type = 1;
string name = 2;
Vector version = 3;
repeated int32 block_indexes = 4 [packed = false];
int32 block_size = 5;
}
enum FileDownloadProgressUpdateType {
FILE_DOWNLOAD_PROGRESS_UPDATE_TYPE_APPEND = 0;
FILE_DOWNLOAD_PROGRESS_UPDATE_TYPE_FORGET = 1;
}
// Ping
message Ping {}
// Close
message Close {
string reason = 1;
}