3 Tor Protocol Specification
8 Note: This is an attempt to specify Tor as it exists as implemented in
9 mid-August, 2004. It is not recommended that others implement this
10 design as it stands; future versions of Tor will implement improved
13 This is not a design document; most design criteria are not examined. For
14 more information on why Tor acts as it does, see tor-design.pdf.
17 - REASON_CONNECTFAILED should include an IP.
18 - Copy prose from tor-design to make everything more readable.
24 K -- a key for a symmetric cypher
26 a|b -- concatenation of 'a' and 'b'.
28 [A0 B1 C2] -- a three-byte sequence, containing the bytes with
29 hexadecimal values A0, B1, and C2, in that order.
31 All numeric values are encoded in network (big-endian) order.
33 Unless otherwise specified, all symmetric ciphers are AES in counter
34 mode, with an IV of all 0 bytes. Asymmetric ciphers are either RSA
35 with 1024-bit keys and exponents of 65537, or DH with the safe prime
36 from rfc2409, section 6.2, whose hex representation is:
38 "FFFFFFFFFFFFFFFFC90FDAA22168C234C4C6628B80DC1CD129024E08"
39 "8A67CC74020BBEA63B139B22514A08798E3404DDEF9519B3CD3A431B"
40 "302B0A6DF25F14374FE1356D6D51C245E485B576625E7EC6F44C42E9"
41 "A637ED6B0BFF5CB6F406B7EDEE386BFB5A899FA5AE9F24117C4B1FE6"
42 "49286651ECE65381FFFFFFFFFFFFFFFF"
44 All "hashes" are 20-byte SHA1 cryptographic digests.
46 When we refer to "the hash of a public key", we mean the SHA1 hash of the
47 ASN.1 encoding of an RSA public key (as specified in PKCS.1).
51 Onion Routing is a distributed overlay network designed to anonymize
52 low-latency TCP-based applications such as web browsing, secure shell,
53 and instant messaging. Clients choose a path through the network and
54 build a ``circuit'', in which each node (or ``onion router'' or ``OR'')
55 in the path knows its predecessor and successor, but no other nodes in
56 the circuit. Traffic flowing down the circuit is sent in fixed-size
57 ``cells'', which are unwrapped by a symmetric key at each node (like
58 the layers of an onion) and relayed downstream.
62 There are two ways to connect to an onion router (OR). The first is
63 as an onion proxy (OP), which allows the OP to authenticate the OR
64 without authenticating itself. The second is as another OR, which
65 allows mutual authentication.
67 Tor uses TLS for link encryption. All implementations MUST support
68 the TLS ciphersuite "TLS_EDH_RSA_WITH_DES_192_CBC3_SHA", and SHOULD
69 support "TLS_DHE_RSA_WITH_AES_128_CBC_SHA" if it is available.
70 Implementations MAY support other ciphersuites, but MUST NOT
71 support any suite without ephemeral keys, symmetric keys of at
72 least 128 bits, and digests of at least 160 bits.
74 An OP or OR always sends a two-certificate chain, consisting of a
75 self-signed certificate containing the OR's identity key, and a second
76 certificate using a short-term connection key. The commonName of the
77 second certificate is the OR's nickname, and the commonName of the first
78 certificate is the OR's nickname, followed by a space and the string
81 All parties receiving certificates must confirm that the identity key is
82 as expected. (When initiating a connection, the expected identity key is
83 the one given in the directory; when creating a connection because of an
84 EXTEND cell, the expected identity key is the one given in the cell.) If
85 the key is not as expected, the party must close the connection.
87 All parties SHOULD reject connections to or from ORs that have malformed
88 or missing certificates. ORs MAY accept connections from OPs with
89 malformed or missing certificates.
91 Once a TLS connection is established, the two sides send cells
92 (specified below) to one another. Cells are sent serially. All
93 cells are 512 bytes long. Cells may be sent embedded in TLS
94 records of any size or divided across TLS records, but the framing
95 of TLS records MUST NOT leak information about the type or contents
98 OR-to-OR connections are never deliberately closed. When an OR
99 starts or receives a new directory, it tries to open new
100 connections to any OR it is not already connected to.
101 [not true, unused OR conns close after 5 mins too -RD]
103 OR-to-OP connections are not permanent. An OP should close a
104 connection to an OR if there are no circuits running over the
105 connection, and an amount of time (KeepalivePeriod, defaults to 5
108 3. Cell Packet format
110 The basic unit of communication for onion routers and onion
111 proxies is a fixed-width "cell". Each cell contains the following
116 Payload (padded with 0 bytes) [509 bytes]
117 [Total size: 512 bytes]
119 The CircID field determines which circuit, if any, the cell is
122 The 'Command' field holds one of the following values:
123 0 -- PADDING (Padding) (See Sec 6.2)
124 1 -- CREATE (Create a circuit) (See Sec 4)
125 2 -- CREATED (Acknowledge create) (See Sec 4)
126 3 -- RELAY (End-to-end data) (See Sec 5)
127 4 -- DESTROY (Stop using a circuit) (See Sec 4)
129 The interpretation of 'Payload' depends on the type of the cell.
130 PADDING: Payload is unused.
131 CREATE: Payload contains the handshake challenge.
132 CREATED: Payload contains the handshake response.
133 RELAY: Payload contains the relay header and relay body.
134 DESTROY: Payload is unused.
135 Upon receiving any other value for the command field, an OR must
138 The payload is padded with 0 bytes.
140 PADDING cells are currently used to implement connection keepalive.
141 If there is no other traffic, ORs and OPs send one another a PADDING
142 cell every few minutes.
144 CREATE, CREATED, and DESTROY cells are used to manage circuits;
147 RELAY cells are used to send commands and data along a circuit; see
150 4. Circuit management
152 4.1. CREATE and CREATED cells
154 Users set up circuits incrementally, one hop at a time. To create a
155 new circuit, OPs send a CREATE cell to the first node, with the
156 first half of the DH handshake; that node responds with a CREATED
157 cell with the second half of the DH handshake plus the first 20 bytes
158 of derivative key data (see section 4.2). To extend a circuit past
159 the first hop, the OP sends an EXTEND relay cell (see section 5)
160 which instructs the last node in the circuit to send a CREATE cell
161 to extend the circuit.
163 The payload for a CREATE cell is an 'onion skin', which consists
164 of the first step of the DH handshake data (also known as g^x).
166 The data is encrypted to Bob's PK as follows: Suppose Bob's PK is
167 L octets long. If the data to be encrypted is shorter than L-42,
168 then it is encrypted directly (with OAEP padding). If the data is at
169 least as long as L-42, then a randomly generated 16-byte symmetric
170 key is prepended to the data, after which the first L-16-42 bytes
171 of the data are encrypted with Bob's PK; and the rest of the data is
172 encrypted with the symmetric key.
174 So in this case, the onion skin on the wire looks like:
176 OAEP padding [42 bytes]
177 Symmetric key [16 bytes]
178 First part of g^x [70 bytes]
179 Symmetrically encrypted:
180 Second part of g^x [58 bytes]
182 The relay payload for an EXTEND relay cell consists of:
185 Onion skin [186 bytes]
186 Public key hash [20 bytes]
188 The port and address field denote the IPV4 address and port of the next
189 onion router in the circuit; the public key hash is the SHA1 hash of the
190 PKCS#1 ASN1 encoding of the next onion router's identity (signing) key.
192 [XXXX Before 0.0.8, EXTEND cells did not include the public key hash.
193 Servers running 0.0.8 distinguish the old-style cells based on the
194 length of payloads. (Servers running 0.0.7 blindly pass on the extend
195 cell regardless of length.) In a future release, old-style EXTEND
196 cells will not be supported.]
198 The payload for a CREATED cell, or the relay payload for an
199 EXTENDED cell, contains:
200 DH data (g^y) [128 bytes]
201 Derivative key data (KH) [20 bytes] <see 4.2 below>
203 The CircID for a CREATE cell is an arbitrarily chosen 2-byte integer,
204 selected by the node (OP or OR) that sends the CREATE cell. To prevent
205 CircID collisions, when one OR sends a CREATE cell to another, it chooses
206 from only one half of the possible values based on the ORs' public
207 identity keys: if the sending OR has a lower key, it chooses a CircID with
208 an MSB of 0; otherwise, it chooses a CircID with an MSB of 1.
210 Public keys are compared numerically by modulus.
212 (Older versions of Tor compared OR nicknames, and did it in a broken and
213 unreliable way. To support versions of Tor earlier than 0.0.9pre6,
214 implementations should notice when the other side of a connection is
215 sending CREATE cells with the "wrong" MSG, and switch accordingly.)
217 4.2. Setting circuit keys
219 Once the handshake between the OP and an OR is completed, both
220 servers can now calculate g^xy with ordinary DH. From the base key
221 material g^xy, they compute derivative key material as follows.
222 First, the server represents g^xy as a big-endian unsigned integer.
223 Next, the server computes 100 bytes of key data as K = SHA1(g^xy |
224 [00]) | SHA1(g^xy | [01]) | ... SHA1(g^xy | [04]) where "00" is
225 a single octet whose value is zero, [01] is a single octet whose
226 value is one, etc. The first 20 bytes of K form KH, bytes 21-40 form
227 the forward digest Df, 41-60 form the backward digest Db, 61-76 form
228 Kf, and 77-92 form Kb.
230 KH is used in the handshake response to demonstrate knowledge of the
231 computed shared key. Df is used to seed the integrity-checking hash
232 for the stream of data going from the OP to the OR, and Db seeds the
233 integrity-checking hash for the data stream from the OR to the OP. Kf
234 is used to encrypt the stream of data going from the OP to the OR, and
235 Kb is used to encrypt the stream of data going from the OR to the OP.
237 4.3. Creating circuits
239 When creating a circuit through the network, the circuit creator
240 (OP) performs the following steps:
242 1. Choose an onion router as an exit node (R_N), such that the onion
243 router's exit policy does not exclude all pending streams
246 2. Choose a chain of (N-1) chain of N onion routers
247 (R_1...R_N-1) to constitute the path, such that no router
248 appears in the path twice.
250 3. If not already connected to the first router in the chain,
251 open a new connection to that router.
253 4. Choose a circID not already in use on the connection with the
254 first router in the chain; send a CREATE cell along the
255 connection, to be received by the first onion router.
257 5. Wait until a CREATED cell is received; finish the handshake
258 and extract the forward key Kf_1 and the backward key Kb_1.
260 6. For each subsequent onion router R (R_2 through R_N), extend
263 To extend the circuit by a single onion router R_M, the OP performs
266 1. Create an onion skin, encrypted to R_M's public key.
268 2. Send the onion skin in a relay EXTEND cell along
269 the circuit (see section 5).
271 3. When a relay EXTENDED cell is received, verify KH, and
272 calculate the shared keys. The circuit is now extended.
274 When an onion router receives an EXTEND relay cell, it sends a CREATE
275 cell to the next onion router, with the enclosed onion skin as its
276 payload. The initiating onion router chooses some circID not yet
277 used on the connection between the two onion routers. (But see
278 section 4.1. above, concerning choosing circIDs based on
279 lexicographic order of nicknames.)
281 As an extension (called router twins), if the desired next onion
282 router R in the circuit is down, and some other onion router R'
283 has the same public keys as R, then it's ok to extend to R' rather than R.
285 When an onion router receives a CREATE cell, if it already has a
286 circuit on the given connection with the given circID, it drops the
287 cell. Otherwise, after receiving the CREATE cell, it completes the
288 DH handshake, and replies with a CREATED cell. Upon receiving a
289 CREATED cell, an onion router packs it payload into an EXTENDED relay
290 cell (see section 5), and sends that cell up the circuit. Upon
291 receiving the EXTENDED relay cell, the OP can retrieve g^y.
293 (As an optimization, OR implementations may delay processing onions
294 until a break in traffic allows time to do so without harming
295 network latency too greatly.)
297 4.4. Tearing down circuits
299 Circuits are torn down when an unrecoverable error occurs along
300 the circuit, or when all streams on a circuit are closed and the
301 circuit's intended lifetime is over. Circuits may be torn down
302 either completely or hop-by-hop.
304 To tear down a circuit completely, an OR or OP sends a DESTROY
305 cell to the adjacent nodes on that circuit, using the appropriate
308 Upon receiving an outgoing DESTROY cell, an OR frees resources
309 associated with the corresponding circuit. If it's not the end of
310 the circuit, it sends a DESTROY cell for that circuit to the next OR
311 in the circuit. If the node is the end of the circuit, then it tears
312 down any associated edge connections (see section 5.1).
314 After a DESTROY cell has been processed, an OR ignores all data or
315 destroy cells for the corresponding circuit.
317 (The rest of this section is not currently used; on errors, circuits
318 are destroyed, not truncated.)
320 To tear down part of a circuit, the OP may send a RELAY_TRUNCATE cell
321 signaling a given OR (Stream ID zero). That OR sends a DESTROY
322 cell to the next node in the circuit, and replies to the OP with a
323 RELAY_TRUNCATED cell.
325 When an unrecoverable error occurs along one connection in a
326 circuit, the nodes on either side of the connection should, if they
327 are able, act as follows: the node closer to the OP should send a
328 RELAY_TRUNCATED cell towards the OP; the node farther from the OP
329 should send a DESTROY cell down the circuit.
331 4.5. Routing relay cells
333 When an OR receives a RELAY cell, it checks the cell's circID and
334 determines whether it has a corresponding circuit along that
335 connection. If not, the OR drops the RELAY cell.
337 Otherwise, if the OR is not at the OP edge of the circuit (that is,
338 either an 'exit node' or a non-edge node), it de/encrypts the payload
339 with AES/CTR, as follows:
340 'Forward' relay cell (same direction as CREATE):
341 Use Kf as key; decrypt.
342 'Back' relay cell (opposite direction from CREATE):
343 Use Kb as key; encrypt.
345 The OR then decides whether it recognizes the relay cell, by
346 inspecting the payload as described in section 5.1 below. If the OR
347 recognizes the cell, it processes the contents of the relay cell.
348 Otherwise, it passes the decrypted relay cell along the circuit if
349 the circuit continues. If the OR at the end of the circuit
350 encounters an unrecognized relay cell, an error has occurred: the OR
351 sends a DESTROY cell to tear down the circuit.
353 When a relay cell arrives at an OP, the OP decrypts the payload
354 with AES/CTR as follows:
355 OP receives data cell:
357 Decrypt with Kb_I. If the payload is recognized (see
358 section 5.1), then stop and process the payload.
360 For more information, see section 5 below.
362 5. Application connections and stream management
366 Within a circuit, the OP and the exit node use the contents of
367 RELAY packets to tunnel end-to-end commands and TCP connections
368 ("Streams") across circuits. End-to-end commands can be initiated
369 by either edge; streams are initiated by the OP.
371 The payload of each unencrypted RELAY cell consists of:
372 Relay command [1 byte]
373 'Recognized' [2 bytes]
379 The relay commands are:
393 The 'Recognized' field in any unencrypted relay payload is always
394 set to zero; the 'digest' field is computed as the first four bytes
395 of the running SHA-1 digest of all the bytes that have travelled
396 over this circuit, seeded from Df or Db respectively (obtained in
397 section 4.2 above), and including this RELAY cell's entire payload
398 (taken with the digest field set to zero).
400 When the 'recognized' field of a RELAY cell is zero, and the digest
401 is correct, the cell is considered "recognized" for the purposes of
402 decryption (see section 4.5 above).
404 All RELAY cells pertaining to the same tunneled stream have the
405 same stream ID. StreamIDs are chosen randomly by the OP. RELAY
406 cells that affect the entire circuit rather than a particular
407 stream use a StreamID of zero.
409 The 'Length' field of a relay cell contains the number of bytes in
410 the relay payload which contain real payload data. The remainder of
411 the payload is padded with NUL bytes.
413 5.2. Opening streams and transferring data
415 To open a new anonymized TCP connection, the OP chooses an open
416 circuit to an exit that may be able to connect to the destination
417 address, selects an arbitrary StreamID not yet used on that circuit,
418 and constructs a RELAY_BEGIN cell with a payload encoding the address
419 and port of the destination host. The payload format is:
421 ADDRESS | ':' | PORT | [00]
423 where ADDRESS can be a DNS hostname, or an IPv4 address in
424 dotted-quad format, or an IPv6 address surrounded by square brackets;
425 and where PORT is encoded in decimal.
427 [What is the [00] for? -NM]
428 [It's so the payload is easy to parse out with string funcs -RD]
430 Upon receiving this cell, the exit node resolves the address as
431 necessary, and opens a new TCP connection to the target port. If the
432 address cannot be resolved, or a connection can't be established, the
433 exit node replies with a RELAY_END cell. (See 5.4 below.)
434 Otherwise, the exit node replies with a RELAY_CONNECTED cell, whose
435 payload is the 4-byte IPv4 address or the 16-byte IPv6 address to which
436 the connection was made.
438 The OP waits for a RELAY_CONNECTED cell before sending any data.
439 Once a connection has been established, the OP and exit node
440 package stream data in RELAY_DATA cells, and upon receiving such
441 cells, echo their contents to the corresponding TCP stream.
442 RELAY_DATA cells sent to unrecognized streams are dropped.
444 Relay RELAY_DROP cells are long-range dummies; upon receiving such
445 a cell, the OR or OP must drop it.
449 When an anonymized TCP connection is closed, or an edge node
450 encounters error on any stream, it sends a 'RELAY_END' cell along the
451 circuit (if possible) and closes the TCP connection immediately. If
452 an edge node receives a 'RELAY_END' cell for any stream, it closes
453 the TCP connection completely, and sends nothing more along the
454 circuit for that stream.
456 The payload of a RELAY_END cell begins with a single 'reason' byte to
457 describe why the stream is closing, plus optional data (depending on
458 the reason.) The values are:
460 1 -- REASON_MISC (catch-all for unlisted reasons)
461 2 -- REASON_RESOLVEFAILED (couldn't look up hostname)
462 3 -- REASON_CONNECTREFUSED (remote host refused connection) [*]
463 4 -- REASON_EXITPOLICY (OR refuses to connect to host or port)
464 5 -- REASON_DESTROY (Circuit is being destroyed)
465 6 -- REASON_DONE (Anonymized TCP connection was closed)
466 7 -- REASON_TIMEOUT (Connection timed out, or OR timed out
468 8 -- (unallocated) [**]
469 9 -- REASON_HIBERNATING (OR is temporarily hibernating)
470 10 -- REASON_INTERNAL (Internal error at the OR)
471 11 -- REASON_RESOURCELIMIT (OR has no resources to fulfill request)
472 12 -- REASON_CONNRESET (Connection was unexpectedly reset)
473 13 -- REASON_TORPROTOCOL (Sent when closing connection because of
474 Tor protocol violations.)
476 (With REASON_EXITPOLICY, the 4-byte IPv4 address or 16-byte IPv6 address
477 forms the optional data; no other reason currently has extra data.)
479 OPs and ORs MUST accept reasons not on the above list, since future
480 versions of Tor may provide more fine-grained reasons.
482 [*] Older versions of Tor also send this reason when connections are
484 [**] Due to a bug in versions of Tor through 0095, error reason 8 must
485 remain allocated until that version is obsolete.
487 --- [The rest of this section describes unimplemented functionality.]
489 Because TCP connections can be half-open, we follow an equivalent
490 to TCP's FIN/FIN-ACK/ACK protocol to close streams.
492 An exit connection can have a TCP stream in one of three states:
493 'OPEN', 'DONE_PACKAGING', and 'DONE_DELIVERING'. For the purposes
494 of modeling transitions, we treat 'CLOSED' as a fourth state,
495 although connections in this state are not, in fact, tracked by the
498 A stream begins in the 'OPEN' state. Upon receiving a 'FIN' from
499 the corresponding TCP connection, the edge node sends a 'RELAY_FIN'
500 cell along the circuit and changes its state to 'DONE_PACKAGING'.
501 Upon receiving a 'RELAY_FIN' cell, an edge node sends a 'FIN' to
502 the corresponding TCP connection (e.g., by calling
503 shutdown(SHUT_WR)) and changing its state to 'DONE_DELIVERING'.
505 When a stream in already in 'DONE_DELIVERING' receives a 'FIN', it
506 also sends a 'RELAY_FIN' along the circuit, and changes its state
507 to 'CLOSED'. When a stream already in 'DONE_PACKAGING' receives a
508 'RELAY_FIN' cell, it sends a 'FIN' and changes its state to
511 If an edge node encounters an error on any stream, it sends a
512 'RELAY_END' cell (if possible) and closes the stream immediately.
514 5.4. Remote hostname lookup
516 To find the address associated with a hostname, the OP sends a
517 RELAY_RESOLVE cell containing the hostname to be resolved. (For a reverse
518 lookup, the OP sends a RELAY_RESOLVE cell containing an in-addr.arpa
519 address.) The OR replies with a RELAY_RESOLVED cell containing a status
520 byte, and any number of answers. Each answer is of the form:
523 Value (variable-width)
524 "Length" is the length of the Value field.
529 0xF0 -- Error, transient
530 0xF1 -- Error, nontransient
532 If any answer has a type of 'Error', then no other answer may be given.
534 The RELAY_RESOLVE cell must use a nonzero, distinct streamID; the
535 corresponding RELAY_RESOLVED cell must use the same streamID. No stream
536 is actually created by the OR when resolving the name.
542 Each node should do appropriate bandwidth throttling to keep its
545 Communicants rely on TCP's default flow control to push back when they
550 Currently nodes are not required to do any sort of link padding or
551 dummy traffic. Because strong attacks exist even with link padding,
552 and because link padding greatly increases the bandwidth requirements
553 for running a node, we plan to leave out link padding until this
554 tradeoff is better understood.
556 6.3. Circuit-level flow control
558 To control a circuit's bandwidth usage, each OR keeps track of
559 two 'windows', consisting of how many RELAY_DATA cells it is
560 allowed to package for transmission, and how many RELAY_DATA cells
561 it is willing to deliver to streams outside the network.
562 Each 'window' value is initially set to 1000 data cells
563 in each direction (cells that are not data cells do not affect
564 the window). When an OR is willing to deliver more cells, it sends a
565 RELAY_SENDME cell towards the OP, with Stream ID zero. When an OR
566 receives a RELAY_SENDME cell with stream ID zero, it increments its
569 Each of these cells increments the corresponding window by 100.
571 The OP behaves identically, except that it must track a packaging
572 window and a delivery window for every OR in the circuit.
574 An OR or OP sends cells to increment its delivery window when the
575 corresponding window value falls under some threshold (900).
577 If a packaging window reaches 0, the OR or OP stops reading from
578 TCP connections for all streams on the corresponding circuit, and
579 sends no more RELAY_DATA cells until receiving a RELAY_SENDME cell.
580 [this stuff is badly worded; copy in the tor-design section -RD]
582 6.4. Stream-level flow control
584 Edge nodes use RELAY_SENDME cells to implement end-to-end flow
585 control for individual connections across circuits. Similarly to
586 circuit-level flow control, edge nodes begin with a window of cells
587 (500) per stream, and increment the window by a fixed value (50)
588 upon receiving a RELAY_SENDME cell. Edge nodes initiate RELAY_SENDME
589 cells when both a) the window is <= 450, and b) there are less than
590 ten cell payloads remaining to be flushed at that edge.
592 7. Directories and routers
594 7.1. Extensible information format
596 Router descriptors and directories both obey the following lightweight
597 extensible information format.
599 The highest level object is a Document, which consists of one or more Items.
600 Every Item begins with a KeywordLine, followed by one or more Objects. A
601 KeywordLine begins with a Keyword, optionally followed by a space and more
602 non-newline characters, and ends with a newline. A Keyword is a sequence of
603 one or more characters in the set [A-Za-z0-9-]. An Object is a block of
604 encoded data in pseudo-Open-PGP-style armor. (cf. RFC 2440)
608 Document ::= (Item | NL)+
609 Item ::= KeywordLine Object*
610 KeywordLine ::= Keyword NL | Keyword SP ArgumentsChar+ NL
611 Keyword = KeywordChar+
612 KeywordChar ::= 'A' ... 'Z' | 'a' ... 'z' | '0' ... '9' | '-'
613 ArgumentChar ::= any printing ASCII character except NL.
614 Object ::= BeginLine Base-64-encoded-data EndLine
615 BeginLine ::= "-----BEGIN " Keyword "-----" NL
616 EndLine ::= "-----END " Keyword "-----" NL
618 The BeginLine and EndLine of an Object must use the same keyword.
620 When interpreting a Document, software MUST reject any document containing a
621 KeywordLine that starts with a keyword it doesn't recognize.
623 The "opt" keyword is reserved for non-critical future extensions. All
624 implementations MUST ignore any item of the form "opt keyword ....." when
625 they would not recognize "keyword ....."; and MUST treat "opt keyword ....."
626 as synonymous with "keyword ......" when keyword is recognized.
628 7.1. Router descriptor format.
630 Every router descriptor MUST start with a "router" Item; MUST end with a
631 "router-signature" Item and an extra NL; and MUST contain exactly one
632 instance of each of the following Items: "published" "onion-key" "link-key"
633 "signing-key" "bandwidth". Additionally, a router descriptor MAY contain any
634 number of "accept", "reject", "fingerprint", "uptime", and "opt" Items.
635 Other than "router" and "router-signature", the items may appear in any
638 The items' formats are as follows:
639 "router" nickname address (ORPort SocksPort DirPort)?
641 Indicates the beginning of a router descriptor. "address" must be an
642 IPv4 address in dotted-quad format. The Port values will soon be
643 deprecated; using them here is equivalent to using them in a "ports"
646 "ports" ORPort SocksPort DirPort
648 Indicates the TCP ports at which this OR exposes functionality.
649 ORPort is a port at which this OR accepts TLS connections for the main
650 OR protocol; SocksPort is the port at which this OR accepts SOCKS
651 connections; and DirPort is the port at which this OR accepts
652 directory-related HTTP connections. If any port is not supported, the
653 value 0 is given instead of a port number.
655 "bandwidth" bandwidth-avg bandwidth-burst bandwidth-observed
657 Estimated bandwidth for this router, in bytes per second. The
658 "average" bandwidth is the volume per second that the OR is willing
659 to sustain over long periods; the "burst" bandwidth is the volume
660 that the OR is willing to sustain in very short intervals. The
661 "observed" value is an estimate of the capacity this server can
662 handle. The server remembers the max bandwidth sustained output
663 over any ten second period in the past day, and another sustained
664 input. The "observed" value is the lesser of these two numbers.
666 [bandwidth-observed was not present before 0.0.8.]
670 A human-readable string describing the system on which this OR is
671 running. This MAY include the operating system, and SHOULD include
672 the name and version of the software implementing the Tor protocol.
674 "published" YYYY-MM-DD HH:MM:SS
676 The time, in GMT, when this descriptor was generated.
680 A fingerprint (20 byte SHA1 hash of asn1 encoded public key, encoded
681 in hex, with spaces after every 4 characters) for this router's
686 The number of seconds that this OR process has been running.
688 "onion-key" NL a public key in PEM format
690 This key is used to encrypt EXTEND cells for this OR. The key MUST
691 be accepted for at least XXXX hours after any new key is published in
692 a subsequent descriptor.
694 "signing-key" NL a public key in PEM format
696 The OR's long-term identity key.
701 These lines, in order, describe the rules that an OR follows when
702 deciding whether to allow a new stream to a given address. The
703 'exitpattern' syntax is described below.
705 "router-signature" NL Signature NL
707 The "SIGNATURE" object contains a signature of the PKCS1-padded SHA1
708 hash of the entire router descriptor, taken from the beginning of the
709 "router" line, through the newline after the "router-signature" line.
710 The router descriptor is invalid unless the signature is performed
711 with the router's identity key.
713 "dircacheport" port NL
715 Same as declaring "port" as this OR's directory port in the 'router'
716 line. At most one of dircacheport and the directory port in the router
717 line may be non-zero.
719 [Obsolete; will go away once 0.0.8 is dead. Older versions of Tor
720 did poorly when non-authoritative directories had a non-zero directory
721 port. To transition, Tor 0.0.8 used dircacheport for
722 nonauthoritative directories.]
726 Describes a way to contact the server's administrator, preferably
727 including an email address and a PGP key fingerprint.
731 'Names' is a space-separated list of server nicknames. If two ORs
732 list one another in their "family" entries, then OPs should treat
733 them as a single OR for the purpose of path selection.
735 For example, if node A's descriptor contains "family B", and node B's
736 descriptor contains "family A", then node A and node B should never
737 be used on the same circuit.
739 "read-history" YYYY-MM-DD HH:MM:SS (NSEC s) NUM,NUM,NUM,NUM,NUM... NL
740 "write-history" YYYY-MM-DD HH:MM:SS (NSEC s) NUM,NUM,NUM,NUM,NUM... NL
742 Declare how much bandwidth the OR has used recently. Usage is divided
743 into intervals of NSEC seconds. The YYYY-MM-DD HH:MM:SS field defines
744 the end of the most recent interval. The numbers are the number of
745 bytes used in the most recent intervals, ordered from oldest to newest.
747 nickname ::= between 1 and 19 alphanumeric characters, case-insensitive.
749 exitpattern ::= addrspec ":" portspec
750 portspec ::= "*" | port | port "-" port
751 port ::= an integer between 1 and 65535, inclusive.
752 addrspec ::= "*" | ip4spec | ip6spec
753 ipv4spec ::= ip4 | ip4 "/" num_ip4_bits | ip4 "/" ip4mask
754 ip4 ::= an IPv4 address in dotted-quad format
755 ip4mask ::= an IPv4 mask in dotted-quad format
756 num_ip4_bits ::= an integer between 0 and 32
757 ip6spec ::= ip6 | ip6 "/" num_ip6_bits
758 ip6 ::= an IPv6 address, surrounded by square brackets.
759 num_ip6_bits ::= an integer between 0 and 128
761 Ports are required; if they are not included in the router
762 line, they must appear in the "ports" lines.
764 7.2. Directory format
766 A Directory begins with a "signed-directory" item, followed by one each of
767 the following, in any order: "recommended-software", "published",
768 "router-status", "directory-signing-key". It may include any number of "opt"
769 items. After these items, a directory includes any number of router
770 descriptors, and a single "directory-signature" item.
774 Indicates the start of a directory.
776 "published" YYYY-MM-DD HH:MM:SS
778 The time at which this directory was generated and signed, in GMT.
780 "directory-signing-key"
782 The key used to sign this directory; see "signing-key" for format.
784 "recommended-software" comma-separated-version-list
786 A list of which versions of which implementations are currently
787 believed to be secure and compatible with the network.
789 "running-routers" space-separated-list
791 A description of which routers are currently believed to be up or
792 down. Every entry consists of an optional "!", followed by either an
793 OR's nickname, or "$" followed by a hexadecimal encoding of the hash
794 of an OR's identity key. If the "!" is included, the router is
795 believed not to be running; otherwise, it is believed to be running.
796 If a router's nickname is given, exactly one router of that nickname
797 will appear in the directory, and that router is "approved" by the
798 directory server. If a hashed identity key is given, that OR is not
799 "approved". [XXXX The 'running-routers' line is only provided for
800 backward compatibility. New code should parse 'router-status'
803 "router-status" space-separated-list
805 A description of which routers are currently believed to be up or
806 down, and which are verified or unverified. Contains one entry for
807 every router that the directory server knows. Each entry is of the
810 !name=$digest [Verified router, currently not live.]
811 name=$digest [Verified router, currently live.]
812 !$digest [Unverified router, currently not live.]
813 or $digest [Unverified router, currently live.]
815 (where 'name' is the router's nickname and 'digest' is a hexadecimal
816 encoding of the hash of the routers' identity key).
818 When parsing this line, clients should only mark a router as
819 'verified' if its nickname AND digest match the one provided.
820 [XXXX 'router-status' was added in 0.0.9pre5; older directory code
821 uses 'running-routers' instead.]
823 "directory-signature" nickname-of-dirserver NL Signature
825 Note: The router descriptor for the directory server MUST appear first.
826 The signature is computed by computing the SHA-1 hash of the
827 directory, from the characters "signed-directory", through the newline
828 after "directory-signature". This digest is then padded with PKCS.1,
829 and signed with the directory server's signing key.
831 If software encounters an unrecognized keyword in a single router descriptor,
832 it should reject only that router descriptor, and continue using the
833 others. If it encounters an unrecognized keyword in the directory header,
834 it should reject the entire directory.
836 7.3. Network-status descriptor
838 A "network-status" (a.k.a "running-routers") document is a truncated
839 directory that contains only the current status of a list of nodes, not
840 their actual descriptors. It contains exactly one of each of the following
847 "published" YYYY-MM-DD HH:MM:SS
855 "directory-signature" NL signature
859 7.4. Behavior of a directory server
861 lists nodes that are connected currently
862 speaks HTTP on a socket, spits out directory on request
864 Directory servers listen on a certain port (the DirPort), and speak a
865 limited version of HTTP 1.0. Clients send either GET or POST commands.
866 The basic interactions are:
867 "%s %s HTTP/1.0\r\nContent-Length: %lu\r\nHost: %s\r\n\r\n",
868 command, url, content-length, host.
869 Get "/tor/" to fetch a full directory.
870 Get "/tor/dir.z" to fetch a compressed full directory.
871 Get "/tor/running-routers" to fetch a network-status descriptor.
872 Post "/tor/" to post a server descriptor, with the body of the
873 request containing the descriptor.
875 "host" is used to specify the address:port of the dirserver, so
876 the request can survive going through HTTP proxies.
878 A.1. Differences between spec and implementation
880 - The current specification requires all ORs to have IPv4 addresses, but
881 allows servers to exit and resolve to IPv6 addresses, and to declare IPv6
882 addresses in their exit policies. The current codebase has no IPv6