2 Tor directory protocol, version 3
4 0. Scope and preliminaries
6 This directory protocol is used by Tor version 0.2.0.x-alpha and later.
7 See dir-spec-v1.txt for information on the protocol used up to the
8 0.1.0.x series, and dir-spec-v2.txt for information on the protocol
9 used by the 0.1.1.x and 0.1.2.x series.
11 Caches and authorities must still support older versions of the
12 directory protocols, until the versions of Tor that require them are
13 finally out of commission.
15 This document merges and supersedes the following proposals:
17 101 Voting on the Tor Directory System
18 103 Splitting identity key from regularly used signing key
19 104 Long and Short Router Descriptors
21 XXX when to download certificates.
25 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL
26 NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and
27 "OPTIONAL" in this document are to be interpreted as described in
32 The earliest versions of Onion Routing shipped with a list of known
33 routers and their keys. When the set of routers changed, users needed to
36 The Version 1 Directory protocol
37 --------------------------------
39 Early versions of Tor (0.0.2) introduced "Directory authorities": servers
40 that served signed "directory" documents containing a list of signed
41 "router descriptors", along with short summary of the status of each
42 router. Thus, clients could get up-to-date information on the state of
43 the network automatically, and be certain that the list they were getting
44 was attested by a trusted directory authority.
46 Later versions (0.0.8) added directory caches, which download
47 directories from the authorities and serve them to clients. Non-caches
48 fetch from the caches in preference to fetching from the authorities, thus
49 distributing bandwidth requirements.
51 Also added during the version 1 directory protocol were "router status"
52 documents: short documents that listed only the up/down status of the
53 routers on the network, rather than a complete list of all the
54 descriptors. Clients and caches would fetch these documents far more
55 frequently than they would fetch full directories.
57 The Version 2 Directory Protocol
58 --------------------------------
60 During the Tor 0.1.1.x series, Tor revised its handling of directory
61 documents in order to address two major problems:
63 * Directories had grown quite large (over 1MB), and most directory
64 downloads consisted mainly of router descriptors that clients
67 * Every directory authority was a trust bottleneck: if a single
68 directory authority lied, it could make clients believe for a time
69 an arbitrarily distorted view of the Tor network. (Clients
70 trusted the most recent signed document they downloaded.) Thus,
71 adding more authorities would make the system less secure, not
74 To address these, we extended the directory protocol so that
75 authorities now published signed "network status" documents. Each
76 network status listed, for every router in the network: a hash of its
77 identity key, a hash of its most recent descriptor, and a summary of
78 what the authority believed about its status. Clients would download
79 the authorities' network status documents in turn, and believe
80 statements about routers iff they were attested to by more than half of
83 Instead of downloading all router descriptors at once, clients
84 downloaded only the descriptors that they did not have. Descriptors
85 were indexed by their digests, in order to prevent malicious caches
86 from giving different versions of a router descriptor to different
89 Routers began working harder to upload new descriptors only when their
90 contents were substantially changed.
93 0.2. Goals of the version 3 protocol
95 Version 3 of the Tor directory protocol tries to solve the following
98 * A great deal of bandwidth used to transmit router descriptors was
99 used by two fields that are not actually used by Tor routers
100 (namely read-history and write-history). We save about 60% by
101 moving them into a separate document that most clients do not
104 * It was possible under certain perverse circumstances for clients
105 to download an unusual set of network status documents, thus
106 partitioning themselves from clients who have a more recent and/or
107 typical set of documents. Even under the best of circumstances,
108 clients were sensitive to the ages of the network status documents
109 they downloaded. Therefore, instead of having the clients
110 correlate multiple network status documents, we have the
111 authorities collectively vote on a single consensus network status
114 * The most sensitive data in the entire network (the identity keys
115 of the directory authorities) needed to be stored unencrypted so
116 that the authorities can sign network-status documents on the fly.
117 Now, the authorities' identity keys are stored offline, and used
118 to certify medium-term signing keys that can be rotated.
120 0.3. Some Remaining questions
122 Things we could solve on a v3 timeframe:
124 The SHA-1 hash is showing its age. We should do something about our
125 dependency on it. We could probably future-proof ourselves here in
126 this revision, at least so far as documents from the authorities are
129 Too many things about the authorities are hardcoded by IP.
131 Perhaps we should start accepting longer identity keys for routers
134 Things to solve eventually:
136 Requiring every client to know about every router won't scale forever.
138 Requiring every directory cache to know every router won't scale
144 There is a small set (say, around 5-10) of semi-trusted directory
145 authorities. A default list of authorities is shipped with the Tor
146 software. Users can change this list, but are encouraged not to do so,
147 in order to avoid partitioning attacks.
149 Every authority has a very-secret, long-term "Authority Identity Key".
150 This is stored encrypted and/or offline, and is used to sign "key
151 certificate" documents. Every key certificate contains a medium-term
152 (3-12 months) "authority signing key", that is used by the authority to
153 sign other directory information. (Note that the authority identity
154 key is distinct from the router identity key that the authority uses
155 in its role as an ordinary router.)
157 Routers periodically upload signed "routers descriptors" to the
158 directory authorities describing their keys, capabilities, and other
159 information. Routers may also upload signed "extra info documents"
160 containing information that is not required for the Tor protocol.
161 Directory authorities serve router descriptors indexed by router
162 identity, or by hash of the descriptor.
164 Routers may act as directory caches to reduce load on the directory
165 authorities. They announce this in their descriptors.
167 Periodically, each directory authority generates a view of
168 the current descriptors and status for known routers. They send a
169 signed summary of this view (a "status vote") to the other
170 authorities. The authorities compute the result of this vote, and sign
171 a "consensus status" document containing the result of the vote.
173 Directory caches download, cache, and re-serve consensus documents.
175 Clients, directory caches, and directory authorities all use consensus
176 documents to find out when their list of routers is out-of-date.
177 (Directory authorities also use vote statuses.) If it is, they download
178 any missing router descriptors. Clients download missing descriptors
179 from caches; caches and authorities download from authorities.
180 Descriptors are downloaded by the hash of the descriptor, not by the
181 server's identity key: this prevents servers from attacking clients by
182 giving them descriptors nobody else uses.
184 All directory information is uploaded and downloaded with HTTP.
186 [Authorities also generate and caches also cache documents produced and
187 used by earlier versions of this protocol; see dir-spec-v1.txt and
188 dir-spec-v2.txt for notes on those versions.]
190 1.1. What's different from version 2?
192 Clients used to download multiple network status documents,
193 corresponding roughly to "status votes" above. They would compute the
194 result of the vote on the client side.
196 Authorities used to sign documents using the same private keys they used
197 for their roles as routers. This forced them to keep these extremely
198 sensitive keys in memory unencrypted.
200 All of the information in extra-info documents used to be kept in the
203 1.2. Document meta-format
205 Router descriptors, directories, and running-routers documents all obey the
206 following lightweight extensible information format.
208 The highest level object is a Document, which consists of one or more
209 Items. Every Item begins with a KeywordLine, followed by zero or more
210 Objects. A KeywordLine begins with a Keyword, optionally followed by
211 whitespace and more non-newline characters, and ends with a newline. A
212 Keyword is a sequence of one or more characters in the set [A-Za-z0-9-].
213 An Object is a block of encoded data in pseudo-Open-PGP-style
214 armor. (cf. RFC 2440)
218 NL = The ascii LF character (hex value 0x0a).
219 Document ::= (Item | NL)+
220 Item ::= KeywordLine Object*
221 KeywordLine ::= Keyword NL | Keyword WS ArgumentChar+ NL
222 Keyword = KeywordChar+
223 KeywordChar ::= 'A' ... 'Z' | 'a' ... 'z' | '0' ... '9' | '-'
224 ArgumentChar ::= any printing ASCII character except NL.
226 Object ::= BeginLine Base-64-encoded-data EndLine
227 BeginLine ::= "-----BEGIN " Keyword "-----" NL
228 EndLine ::= "-----END " Keyword "-----" NL
230 The BeginLine and EndLine of an Object must use the same keyword.
232 When interpreting a Document, software MUST ignore any KeywordLine that
233 starts with a keyword it doesn't recognize; future implementations MUST NOT
234 require current clients to understand any KeywordLine not currently
237 The "opt" keyword was used until Tor 0.1.2.5-alpha for non-critical future
238 extensions. All implementations MUST ignore any item of the form "opt
239 keyword ....." when they would not recognize "keyword ....."; and MUST
240 treat "opt keyword ....." as synonymous with "keyword ......" when keyword
243 Implementations before 0.1.2.5-alpha rejected any document with a
244 KeywordLine that started with a keyword that they didn't recognize.
245 When generating documents that need to be read by older versions of Tor,
246 implementations MUST prefix items not recognized by older versions of
247 Tor with an "opt" until those versions of Tor are obsolete. [Note that
248 key certificates, status vote documents, extra info documents, and
249 status consensus documents will never be read by older versions of Tor.]
251 Other implementations that want to extend Tor's directory format MAY
252 introduce their own items. The keywords for extension items SHOULD start
253 with the characters "x-" or "X-", to guarantee that they will not conflict
254 with keywords used by future versions of Tor.
256 In our document descriptions below, we tag Items with a multiplicity in
257 brackets. Possible tags are:
259 "At start, exactly once": These items MUST occur in every instance of
260 the document type, and MUST appear exactly once, and MUST be the
261 first item in their documents.
263 "Exactly once": These items MUST occur exactly one time in every
264 instance of the document type.
266 "At end, exactly once": These items MUST occur in every instance of
267 the document type, and MUST appear exactly once, and MUST be the
268 last item in their documents.
270 "At most once": These items MAY occur zero or one times in any
271 instance of the document type, but MUST NOT occur more than once.
273 "Any number": These items MAY occur zero, one, or more times in any
274 instance of the document type.
276 "Once or more": These items MUST occur at least once in any instance
277 of the document type, and MAY occur more.
279 1.3. Signing documents
281 Every signable document below is signed in a similar manner, using a
282 given "Initial Item", a final "Signature Item", a digest algorithm, and
285 The Initial Item must be the first item in the document.
287 The Signature Item has the following format:
289 <signature item keyword> [arguments] NL SIGNATURE NL
291 The "SIGNATURE" Object contains a signature (using the signing key) of
292 the PKCS1-padded digest of the entire document, taken from the
293 beginning of the Initial item, through the newline after the Signature
294 Item's keyword and its arguments.
296 Unless otherwise, the digest algorithm is SHA-1.
298 All documents are invalid unless signed with the correct signing key.
300 The "Digest" of a document, unless stated otherwise, is its digest *as
301 signed by this signature scheme*.
305 Every consensus document has a "valid-after" (VA) time, a "fresh-until"
306 (FU) time and a "valid-until" (VU) time. VA MUST precede FU, which MUST
307 in turn precede VU. Times are chosen so that every consensus will be
308 "fresh" until the next consensus becomes valid, and "valid" for a while
309 after. At least 3 consensuses should be valid at any given time.
311 The timeline for a given consensus is as follows:
313 VA-DistSeconds-VoteSeconds: The authorities exchange votes.
315 VA-DistSeconds-VoteSeconds/2: The authorities try to download any
316 votes they don't have.
318 VA-DistSeconds: The authorities calculate the consensus and exchange
321 VA-DistSeconds/2: The authorities try to download any signatures
324 VA: All authorities have a multiply signed consensus.
326 VA ... FU: Caches download the consensus. (Note that since caches have
327 no way of telling what VA and FU are until they have downloaded
328 the consensus, they assume that the present consensus's VA is
329 equal to the previous one's FU, and that its FU is one interval after
332 FU: The consensus is no longer the freshest consensus.
334 FU ... (the current consensus's VU): Clients download the consensus.
335 (See note above: clients guess that the next consensus's FU will be
336 two intervals after the current VA.)
338 VU: The consensus is no longer valid.
340 VoteSeconds and DistSeconds MUST each be at least 20 seconds; FU-VA and
341 VU-FU MUST each be at least 5 minutes.
343 2. Router operation and formats
345 ORs SHOULD generate a new router descriptor and a new extra-info
346 document whenever any of the following events have occurred:
348 - A period of time (18 hrs by default) has passed since the last
349 time a descriptor was generated.
351 - A descriptor field other than bandwidth or uptime has changed.
353 - Bandwidth has changed by a factor of 2 from the last time a
354 descriptor was generated, and at least a given interval of time
355 (20 mins by default) has passed since then.
357 - Its uptime has been reset (by restarting).
359 [XXX this list is incomplete; see router_differences_are_cosmetic()
360 in routerlist.c for others]
362 ORs SHOULD NOT publish a new router descriptor or extra-info document
363 if none of the above events have occurred and not much time has passed
364 (12 hours by default).
366 After generating a descriptor, ORs upload them to every directory
367 authority they know, by posting them (in order) to the URL
369 http://<hostname:port>/tor/
371 2.1. Router descriptor format
373 Router descriptors consist of the following items. For backward
374 compatibility, there should be an extra NL at the end of each router
377 In lines that take multiple arguments, extra arguments SHOULD be
378 accepted and ignored. Many of the nonterminals below are defined in
381 "router" nickname address ORPort SOCKSPort DirPort NL
383 [At start, exactly once.]
385 Indicates the beginning of a router descriptor. "nickname" must be a
386 valid router nickname as specified in 2.3. "address" must be an IPv4
387 address in dotted-quad format. The last three numbers indicate the
388 TCP ports at which this OR exposes functionality. ORPort is a port at
389 which this OR accepts TLS connections for the main OR protocol;
390 SOCKSPort is deprecated and should always be 0; and DirPort is the
391 port at which this OR accepts directory-related HTTP connections. If
392 any port is not supported, the value 0 is given instead of a port
393 number. (At least one of DirPort and ORPort SHOULD be set;
394 authorities MAY reject any descriptor with both DirPort and ORPort of
397 "bandwidth" bandwidth-avg bandwidth-burst bandwidth-observed NL
401 Estimated bandwidth for this router, in bytes per second. The
402 "average" bandwidth is the volume per second that the OR is willing to
403 sustain over long periods; the "burst" bandwidth is the volume that
404 the OR is willing to sustain in very short intervals. The "observed"
405 value is an estimate of the capacity this server can handle. The
406 server remembers the max bandwidth sustained output over any ten
407 second period in the past day, and another sustained input. The
408 "observed" value is the lesser of these two numbers.
414 A human-readable string describing the system on which this OR is
415 running. This MAY include the operating system, and SHOULD include
416 the name and version of the software implementing the Tor protocol.
418 "published" YYYY-MM-DD HH:MM:SS NL
422 The time, in GMT, when this descriptor (and its corresponding
423 extra-info document if any) was generated.
425 "fingerprint" fingerprint NL
429 A fingerprint (a HASH_LEN-byte of asn1 encoded public key, encoded in
430 hex, with a single space after every 4 characters) for this router's
431 identity key. A descriptor is considered invalid (and MUST be
432 rejected) if the fingerprint line does not match the public key.
434 [We didn't start parsing this line until Tor 0.1.0.6-rc; it should
435 be marked with "opt" until earlier versions of Tor are obsolete.]
437 "hibernating" bool NL
441 If the value is 1, then the Tor server was hibernating when the
442 descriptor was published, and shouldn't be used to build circuits.
444 [We didn't start parsing this line until Tor 0.1.0.6-rc; it should be
445 marked with "opt" until earlier versions of Tor are obsolete.]
451 The number of seconds that this OR process has been running.
453 "onion-key" NL a public key in PEM format
457 This key is used to encrypt EXTEND cells for this OR. The key MUST be
458 accepted for at least 1 week after any new key is published in a
459 subsequent descriptor. It MUST be 1024 bits.
461 "signing-key" NL a public key in PEM format
465 The OR's long-term identity key. It MUST be 1024 bits.
467 "accept" exitpattern NL
468 "reject" exitpattern NL
472 These lines describe an "exit policy": the rules that an OR follows
473 when deciding whether to allow a new stream to a given address. The
474 'exitpattern' syntax is described below. There MUST be at least one
475 such entry. The rules are considered in order; if no rule matches,
476 the address will be accepted. For clarity, the last such entry SHOULD
477 be accept *:* or reject *:*.
479 "router-signature" NL Signature NL
481 [At end, exactly once]
483 The "SIGNATURE" object contains a signature of the PKCS1-padded
484 hash of the entire router descriptor, taken from the beginning of the
485 "router" line, through the newline after the "router-signature" line.
486 The router descriptor is invalid unless the signature is performed
487 with the router's identity key.
493 Describes a way to contact the server's administrator, preferably
494 including an email address and a PGP key fingerprint.
500 'Names' is a space-separated list of server nicknames or
501 hexdigests. If two ORs list one another in their "family" entries,
502 then OPs should treat them as a single OR for the purpose of path
505 For example, if node A's descriptor contains "family B", and node B's
506 descriptor contains "family A", then node A and node B should never
507 be used on the same circuit.
509 "read-history" YYYY-MM-DD HH:MM:SS (NSEC s) NUM,NUM,NUM,NUM,NUM... NL
511 "write-history" YYYY-MM-DD HH:MM:SS (NSEC s) NUM,NUM,NUM,NUM,NUM... NL
514 Declare how much bandwidth the OR has used recently. Usage is divided
515 into intervals of NSEC seconds. The YYYY-MM-DD HH:MM:SS field
516 defines the end of the most recent interval. The numbers are the
517 number of bytes used in the most recent intervals, ordered from
520 [We didn't start parsing these lines until Tor 0.1.0.6-rc; they should
521 be marked with "opt" until earlier versions of Tor are obsolete.]
523 [See also migration notes in section 2.2.1.]
529 Declare whether this version of Tor is using the newer enhanced
530 dns logic. Versions of Tor with this field set to false SHOULD NOT
531 be used for reverse hostname lookups.
533 [This option is obsolete. All Tor current servers should be presumed
534 to have the evdns backend.]
536 "caches-extra-info" NL
540 Present only if this router is a directory cache that provides
541 extra-info documents.
543 [Versions before 0.2.0.1-alpha don't recognize this, and versions
544 before 0.1.2.5-alpha will reject descriptors containing it unless
545 it is prefixed with "opt"; it should be so prefixed until these
546 versions are obsolete.]
548 "extra-info-digest" digest NL
552 "Digest" is a hex-encoded digest (using upper-case characters) of the
553 router's extra-info document, as signed in the router's extra-info
554 (that is, not including the signature). (If this field is absent, the
555 router is not uploading a corresponding extra-info document.)
557 [Versions before 0.2.0.1-alpha don't recognize this, and versions
558 before 0.1.2.5-alpha will reject descriptors containing it unless
559 it is prefixed with "opt"; it should be so prefixed until these
560 versions are obsolete.]
562 "hidden-service-dir" *(SP VersionNum) NL
566 Present only if this router stores and serves hidden service
567 descriptors. If any VersionNum(s) are specified, this router
568 supports those descriptor versions. If none are specified, it
569 defaults to version 2 descriptors.
571 [Versions of Tor before 0.1.2.5-alpha rejected router descriptors
572 with unrecognized items; the protocols line should be preceded with
573 an "opt" until these Tors are obsolete.]
575 "protocols" SP "Link" SP LINK-VERSION-LIST SP "Circuit" SP
576 CIRCUIT-VERSION-LIST NL
580 Both lists are space-separated sequences of numbers, to indicate which
581 protocols the server supports. As of 30 Mar 2008, specified
582 protocols are "Link 1 2 Circuit 1". See section 4.1 of tor-spec.txt
583 for more information about link protocol versions.
585 [Versions of Tor before 0.1.2.5-alpha rejected router descriptors
586 with unrecognized items; the protocols line should be preceded with
587 an "opt" until these Tors are obsolete.]
589 "allow-single-hop-exits" NL
593 Present only if the router allows single-hop circuits to make exit
594 connections. Most Tor servers do not support this: this is
595 included for specialized controllers designed to support perspective
599 2.2. Extra-info documents
601 Extra-info documents consist of the following items:
603 "extra-info" Nickname Fingerprint NL
604 [At start, exactly once.]
606 Identifies what router this is an extra info descriptor for.
607 Fingerprint is encoded in hex (using upper-case letters), with
610 "published" YYYY-MM-DD HH:MM:SS NL
614 The time, in GMT, when this document (and its corresponding router
615 descriptor if any) was generated. It MUST match the published time
616 in the corresponding router descriptor.
618 "read-history" YYYY-MM-DD HH:MM:SS (NSEC s) NUM,NUM,NUM,NUM,NUM... NL
620 "write-history" YYYY-MM-DD HH:MM:SS (NSEC s) NUM,NUM,NUM,NUM,NUM... NL
623 As documented in 2.1 above. See migration notes in section 2.2.1.
625 "geoip-db-digest" Digest NL
628 SHA1 digest of the GeoIP database file that is used to resolve IP
629 addresses to country codes.
631 ("geoip-start" YYYY-MM-DD HH:MM:SS NL)
632 ("geoip-client-origins" CC=N,CC=N,... NL)
634 Only generated by bridge routers (see blocking.pdf), and only
635 when they have been configured with a geoip database.
636 Non-bridges SHOULD NOT generate these fields. Contains a list
637 of mappings from two-letter country codes (CC) to the number
638 of clients that have connected to that bridge from that
639 country (approximate, and rounded up to the nearest multiple of 8
640 in order to hamper traffic analysis). A country is included
641 only if it has at least one address. The time in
642 "geoip-start" is the time at which we began collecting geoip
645 "geoip-start" and "geoip-client-origins" have been replaced by
646 "bridge-stats-end" and "bridge-stats-ips" in 0.2.2.4-alpha. The
647 reason is that the measurement interval with "geoip-stats" as
648 determined by subtracting "geoip-start" from "published" could
649 have had a variable length, whereas the measurement interval in
650 0.2.2.4-alpha and later is set to be exactly 24 hours long. In
651 order to clearly distinguish the new measurement intervals from
652 the old ones, the new keywords have been introduced.
654 "bridge-stats-end" YYYY-MM-DD HH:MM:SS (NSEC s) NL
657 YYYY-MM-DD HH:MM:SS defines the end of the included measurement
658 interval of length NSEC seconds (86400 seconds by default).
660 A "bridge-stats-end" line, as well as any other "bridge-*" line,
661 is only added when the relay has been running as a bridge for at
664 "bridge-ips" CC=N,CC=N,... NL
667 List of mappings from two-letter country codes to the number of
668 unique IP addresses that have connected from that country to the
669 bridge and which are no known relays, rounded up to the nearest
672 "dirreq-stats-end" YYYY-MM-DD HH:MM:SS (NSEC s) NL
675 YYYY-MM-DD HH:MM:SS defines the end of the included measurement
676 interval of length NSEC seconds (86400 seconds by default).
678 A "dirreq-stats-end" line, as well as any other "dirreq-*" line,
679 is only added when the relay has opened its Dir port and after 24
680 hours of measuring directory requests.
682 "dirreq-v2-ips" CC=N,CC=N,... NL
684 "dirreq-v3-ips" CC=N,CC=N,... NL
687 List of mappings from two-letter country codes to the number of
688 unique IP addresses that have connected from that country to
689 request a v2/v3 network status, rounded up to the nearest multiple
690 of 8. Only those IP addresses are counted that the directory can
691 answer with a 200 OK status code.
693 "dirreq-v2-reqs" CC=N,CC=N,... NL
695 "dirreq-v3-reqs" CC=N,CC=N,... NL
698 List of mappings from two-letter country codes to the number of
699 requests for v2/v3 network statuses from that country, rounded up
700 to the nearest multiple of 8. Only those requests are counted that
701 the directory can answer with a 200 OK status code.
703 "dirreq-v2-share" num% NL
705 "dirreq-v3-share" num% NL
708 The share of v2/v3 network status requests that the directory
709 expects to receive from clients based on its advertised bandwidth
710 compared to the overall network bandwidth capacity. Shares are
711 formatted in percent with two decimal places. Shares are
712 calculated as means over the whole 24-hour interval.
714 "dirreq-v2-resp" status=num,... NL
716 "dirreq-v3-resp" status=nul,... NL
719 List of mappings from response statuses to the number of requests
720 for v2/v3 network statuses that were answered with that response
721 status, rounded up to the nearest multiple of 4. Only response
722 statuses with at least 1 response are reported. New response
723 statuses can be added at any time. The current list of response
724 statuses is as follows:
726 "ok": a network status request is answered; this number
727 corresponds to the sum of all requests as reported in
728 "dirreq-v2-reqs" or "dirreq-v3-reqs", respectively, before
730 "not-enough-sigs: a version 3 network status is not signed by a
731 sufficient number of requested authorities.
732 "unavailable": a requested network status object is unavailable.
733 "not-found": a requested network status is not found.
734 "not-modified": a network status has not been modified since the
735 If-Modified-Since time that is included in the request.
736 "busy": the directory is busy.
738 "dirreq-v2-direct-dl" key=val,... NL
740 "dirreq-v3-direct-dl" key=val,... NL
742 "dirreq-v2-tunneled-dl" key=val,... NL
744 "dirreq-v3-tunneled-dl" key=val,... NL
747 List of statistics about possible failures in the download process
748 of v2/v3 network statuses. Requests are either "direct"
749 HTTP-encoded requests over the relay's directory port, or
750 "tunneled" requests using a BEGIN_DIR cell over the relay's OR
751 port. The list of possible statistics can change, and statistics
752 can be left out from reporting. The current list of statistics is
755 Successful downloads and failures:
757 "complete": a client has finished the download successfully.
758 "timeout": a download did not finish within 10 minutes after
759 starting to send the response.
760 "running": a download is still running at the end of the
761 measurement period for less than 10 minutes after starting to
766 "min", "max": smallest and largest measured bandwidth in B/s.
767 "d[1-4,6-9]": 1st to 4th and 6th to 9th decile of measured
768 bandwidth in B/s. For a given decile i, i/10 of all downloads
769 had a smaller bandwidth than di, and (10-i)/10 of all downloads
770 had a larger bandwidth than di.
771 "q[1,3]": 1st and 3rd quartile of measured bandwidth in B/s. One
772 fourth of all downloads had a smaller bandwidth than q1, one
773 fourth of all downloads had a larger bandwidth than q3, and the
774 remaining half of all downloads had a bandwidth between q1 and
776 "md": median of measured bandwidth in B/s. Half of the downloads
777 had a smaller bandwidth than md, the other half had a larger
780 "dirreq-read-history" YYYY-MM-DD HH:MM:SS (NSEC s) NUM,NUM,NUM... NL
782 "dirreq-write-history" YYYY-MM-DD HH:MM:SS (NSEC s) NUM,NUM,NUM... NL
785 Declare how much bandwidth the OR has spent on answering directory
786 requests. Usage is divided into intervals of NSEC seconds. The
787 YYYY-MM-DD HH:MM:SS field defines the end of the most recent
788 interval. The numbers are the number of bytes used in the most
789 recent intervals, ordered from oldest to newest.
791 "entry-stats-end" YYYY-MM-DD HH:MM:SS (NSEC s) NL
794 YYYY-MM-DD HH:MM:SS defines the end of the included measurement
795 interval of length NSEC seconds (86400 seconds by default).
797 An "entry-stats-end" line, as well as any other "entry-*"
798 line, is first added after the relay has been running for at least
801 "entry-ips" CC=N,CC=N,... NL
804 List of mappings from two-letter country codes to the number of
805 unique IP addresses that have connected from that country to the
806 relay and which are no known other relays, rounded up to the
807 nearest multiple of 8.
809 "cell-stats-end" YYYY-MM-DD HH:MM:SS (NSEC s) NL
812 YYYY-MM-DD HH:MM:SS defines the end of the included measurement
813 interval of length NSEC seconds (86400 seconds by default).
815 A "cell-stats-end" line, as well as any other "cell-*" line,
816 is first added after the relay has been running for at least 24
819 "cell-processed-cells" num,...,num NL
822 Mean number of processed cells per circuit, subdivided into
823 deciles of circuits by the number of cells they have processed in
824 descending order from loudest to quietest circuits.
826 "cell-queued-cells" num,...,num NL
829 Mean number of cells contained in queues by circuit decile. These
830 means are calculated by 1) determining the mean number of cells in
831 a single circuit between its creation and its termination and 2)
832 calculating the mean for all circuits in a given decile as
833 determined in "cell-processed-cells". Numbers have a precision of
836 "cell-time-in-queue" num,...,num NL
839 Mean time cells spend in circuit queues in milliseconds. Times are
840 calculated by 1) determining the mean time cells spend in the
841 queue of a single circuit and 2) calculating the mean for all
842 circuits in a given decile as determined in
843 "cell-processed-cells".
845 "cell-circuits-per-decile" num NL
848 Mean number of circuits that are included in any of the deciles,
849 rounded up to the next integer.
851 "conn-bi-direct" YYYY-MM-DD HH:MM:SS (NSEC s) BELOW,READ,WRITE,BOTH NL
854 Number of connections, split into 10-second intervals, that are
855 used uni-directionally or bi-directionally as observed in the NSEC
856 seconds (usually 86400 seconds) before YYYY-MM-DD HH:MM:SS. Every
857 10 seconds, we determine for every connection whether we read and
858 wrote less than a threshold of 20 KiB (BELOW), read at least 10
859 times more than we wrote (READ), wrote at least 10 times more than
860 we read (WRITE), or read and wrote more than the threshold, but
861 not 10 times more in either direction (BOTH). After classifying a
862 connection, read and write counters are reset for the next
865 "exit-stats-end" YYYY-MM-DD HH:MM:SS (NSEC s) NL
868 YYYY-MM-DD HH:MM:SS defines the end of the included measurement
869 interval of length NSEC seconds (86400 seconds by default).
871 An "exit-stats-end" line, as well as any other "exit-*" line, is
872 first added after the relay has been running for at least 24 hours
873 and only if the relay permits exiting (where exiting to a single
874 port and IP address is sufficient).
876 "exit-kibibytes-written" port=N,port=N,... NL
878 "exit-kibibytes-read" port=N,port=N,... NL
881 List of mappings from ports to the number of kibibytes that the
882 relay has written to or read from exit connections to that port,
883 rounded up to the next full kibibyte.
885 "exit-streams-opened" port=N,port=N,... NL
888 List of mappings from ports to the number of opened exit streams
889 to that port, rounded up to the nearest multiple of 4.
891 "router-signature" NL Signature NL
892 [At end, exactly once.]
894 A document signature as documented in section 1.3, using the
895 initial item "extra-info" and the final item "router-signature",
896 signed with the router's identity key.
898 2.2.1. Moving history fields to extra-info documents.
900 Tools that want to use the read-history and write-history values SHOULD
901 download extra-info documents as well as router descriptors. Such
902 tools SHOULD accept history values from both sources; if they appear in
903 both documents, the values in the extra-info documents are authoritative.
905 New versions of Tor no longer generate router descriptors
906 containing read-history or write-history. Tools should continue to
907 accept read-history and write-history values in router descriptors
908 produced by older versions of Tor until all Tor versions earlier
909 than 0.2.0.x are obsolete.
911 2.3. Nonterminals in router descriptors
913 nickname ::= between 1 and 19 alphanumeric characters ([A-Za-z0-9]),
915 hexdigest ::= a '$', followed by 40 hexadecimal characters
916 ([A-Fa-f0-9]). [Represents a server by the digest of its identity
919 exitpattern ::= addrspec ":" portspec
920 portspec ::= "*" | port | port "-" port
921 port ::= an integer between 1 and 65535, inclusive.
923 [Some implementations incorrectly generate ports with value 0.
924 Implementations SHOULD accept this, and SHOULD NOT generate it.
925 Connections to port 0 are never permitted.]
927 addrspec ::= "*" | ip4spec | ip6spec
928 ipv4spec ::= ip4 | ip4 "/" num_ip4_bits | ip4 "/" ip4mask
929 ip4 ::= an IPv4 address in dotted-quad format
930 ip4mask ::= an IPv4 mask in dotted-quad format
931 num_ip4_bits ::= an integer between 0 and 32
932 ip6spec ::= ip6 | ip6 "/" num_ip6_bits
933 ip6 ::= an IPv6 address, surrounded by square brackets.
934 num_ip6_bits ::= an integer between 0 and 128
938 3. Formats produced by directory authorities.
940 Every authority has two keys used in this protocol: a signing key, and
941 an authority identity key. (Authorities also have a router identity
942 key used in their role as a router and by earlier versions of the
943 directory protocol.) The identity key is used from time to time to
944 sign new key certificates using new signing keys; it is very sensitive.
945 The signing key is used to sign key certificates and status documents.
947 There are three kinds of documents generated by directory authorities:
953 Each is discussed below.
955 3.1. Key certificates
957 Key certificates consist of the following items:
959 "dir-key-certificate-version" version NL
961 [At start, exactly once.]
963 Determines the version of the key certificate. MUST be "3" for
964 the protocol described in this document. Implementations MUST
965 reject formats they don't understand.
967 "dir-address" IPPort NL
970 An IP:Port for this authority's directory port.
972 "fingerprint" fingerprint NL
976 Hexadecimal encoding without spaces based on the authority's
979 "dir-identity-key" NL a public key in PEM format
983 The long-term authority identity key for this authority. This key
984 SHOULD be at least 2048 bits long; it MUST NOT be shorter than
987 "dir-key-published" YYYY-MM-DD HH:MM:SS NL
991 The time (in GMT) when this document and corresponding key were
994 "dir-key-expires" YYYY-MM-DD HH:MM:SS NL
998 A time (in GMT) after which this key is no longer valid.
1000 "dir-signing-key" NL a key in PEM format
1004 The directory server's public signing key. This key MUST be at
1005 least 1024 bits, and MAY be longer.
1007 "dir-key-crosscert" NL CrossSignature NL
1011 NOTE: Authorities MUST include this field in all newly generated
1012 certificates. A future version of this specification will make
1015 CrossSignature is a signature, made using the certificate's signing
1016 key, of the digest of the PKCS1-padded hash of the certificate's
1017 identity key. For backward compatibility with broken versions of the
1018 parser, we wrap the base64-encoded signature in -----BEGIN ID
1019 SIGNATURE---- and -----END ID SIGNATURE----- tags. Implementations
1020 MUST allow the "ID " portion to be omitted, however.
1022 When encountering a certificate with a dir-key-crosscert entry,
1023 implementations MUST verify that the signature is a correct signature
1024 of the hash of the identity key using the signing key.
1026 "dir-key-certification" NL Signature NL
1028 [At end, exactly once.]
1030 A document signature as documented in section 1.3, using the
1031 initial item "dir-key-certificate-version" and the final item
1032 "dir-key-certification", signed with the authority identity key.
1034 Authorities MUST generate a new signing key and corresponding
1035 certificate before the key expires.
1037 3.2. Vote and consensus status documents
1039 Votes and consensuses are more strictly formatted then other documents
1040 in this specification, since different authorities must be able to
1041 generate exactly the same consensus given the same set of votes.
1043 The procedure for deciding when to generate vote and consensus status
1044 documents are described in section 1.4 on the voting timeline.
1046 Status documents contain a preamble, an authority section, a list of
1047 router status entries, and one or more footer signature, in that order.
1049 Unlike other formats described above, a SP in these documents must be a
1050 single space character (hex 20).
1052 Some items appear only in votes, and some items appear only in
1053 consensuses. Unless specified, items occur in both.
1055 The preamble contains the following items. They MUST occur in the
1058 "network-status-version" SP version NL.
1060 [At start, exactly once.]
1062 A document format version. For this specification, the version is
1065 "vote-status" SP type NL
1069 The status MUST be "vote" or "consensus", depending on the type of
1072 "consensus-methods" SP IntegerList NL
1074 [Exactly once for votes; does not occur in consensuses.]
1076 A space-separated list of supported methods for generating
1077 consensuses from votes. See section 3.4.1 for details. Method "1"
1080 "consensus-method" SP Integer NL
1082 [Exactly once for consensuses; does not occur in votes.]
1084 See section 3.4.1 for details.
1086 (Only included when the vote is generated with consensus-method 2 or
1089 "published" SP YYYY-MM-DD SP HH:MM:SS NL
1091 [Exactly once for votes; does not occur in consensuses.]
1093 The publication time for this status document (if a vote).
1095 "valid-after" SP YYYY-MM-DD SP HH:MM:SS NL
1099 The start of the Interval for this vote. Before this time, the
1100 consensus document produced from this vote should not be used.
1101 See 1.4 for voting timeline information.
1103 "fresh-until" SP YYYY-MM-DD SP HH:MM:SS NL
1107 The time at which the next consensus should be produced; before this
1108 time, there is no point in downloading another consensus, since there
1109 won't be a new one. See 1.4 for voting timeline information.
1111 "valid-until" SP YYYY-MM-DD SP HH:MM:SS NL
1115 The end of the Interval for this vote. After this time, the
1116 consensus produced by this vote should not be used. See 1.4 for
1117 voting timeline information.
1119 "voting-delay" SP VoteSeconds SP DistSeconds NL
1123 VoteSeconds is the number of seconds that we will allow to collect
1124 votes from all authorities; DistSeconds is the number of seconds
1125 we'll allow to collect signatures from all authorities. See 1.4 for
1126 voting timeline information.
1128 "client-versions" SP VersionList NL
1132 A comma-separated list of recommended client versions, in
1133 ascending order. If absent, no opinion is held about client
1136 "server-versions" SP VersionList NL
1140 A comma-separated list of recommended server versions, in
1141 ascending order. If absent, no opinion is held about server
1144 "known-flags" SP FlagList NL
1148 A space-separated list of all of the flags that this document
1149 might contain. A flag is "known" either because the authority
1150 knows about them and might set them (if in a vote), or because
1151 enough votes were counted for the consensus for an authoritative
1152 opinion to have been formed about their status.
1154 "params" SP [Parameters] NL
1158 Parameter ::= Keyword '=' Int32
1159 Int32 ::= A decimal integer between -2147483648 and 2147483647.
1160 Parameters ::= Parameter | Parameters SP Parameter
1162 The parameters list, if present, contains a space-separated list of
1163 case-sensitive key-value pairs, sorted in lexical order by their
1164 keyword (as ASCII byte strings). Each parameter has its own meaning.
1166 (Only included when the vote is generated with consensus-method 7 or
1169 Commonly used "param" arguments at this point include:
1171 "circwindow" -- the default package window that circuits should
1172 be established with. It started out at 1000 cells, but some
1173 research indicates that a lower value would mean fewer cells in
1174 transit in the network at any given time. Obeyed by Tor 0.2.1.20
1178 "CircuitPriorityHalflifeMsec" -- the halflife parameter used when
1179 weighting which circuit will send the next cell. Obeyed by Tor
1180 0.2.2.10-alpha and later. (Versions of Tor between 0.2.2.7-alpha
1181 and 0.2.2.10-alpha recognized a "CircPriorityHalflifeMsec" parameter,
1182 but mishandled it badly.)
1183 Min: -1, Max: 2147483647 (INT32_MAX)
1185 "perconnbwrate" and "perconnbwburst" -- if set, each relay sets
1186 up a separate token bucket for every client OR connection,
1187 and rate limits that connection indepedently. Typically left
1188 unset, except when used for performance experiments around trac
1189 entry 1750. Only honored by relays running Tor 0.2.2.16-alpha
1190 and later. (Note that relays running 0.2.2.7-alpha through
1191 0.2.2.14-alpha looked for bwconnrate and bwconnburst, but then
1192 did the wrong thing with them; see bug 1830 for details.)
1193 Min: 1, Max: 2147483647 (INT32_MAX)
1195 "refuseunknownexits" -- if set to one, exit relays look at
1196 the previous hop of circuits that ask to open an exit stream,
1197 and refuse to exit if they don't recognize it as a relay. The
1198 goal is to make it harder for people to use them as one-hop
1199 proxies. See trac entry 1751 for details.
1202 "cbtdisabled", "cbtnummodes", "cbtrecentcount", "cbtmaxtimeouts",
1203 "cbtmincircs", "cbtquantile", "cbtclosequantile", "cbttestfreq",
1204 "cbtmintimeout", and "cbtinitialtimeout" -- see "2.4.5. Consensus
1205 parameters governing behavior" in path-spec.txt for a series of
1206 circuit build time related consensus params.
1208 The authority section of a vote contains the following items, followed
1209 in turn by the authority's current key certificate:
1211 "dir-source" SP nickname SP identity SP address SP IP SP dirport SP
1214 [Exactly once, at start]
1216 Describes this authority. The nickname is a convenient identifier
1217 for the authority. The identity is an uppercase hex fingerprint of
1218 the authority's current (v3 authority) identity key. The address is
1219 the server's hostname. The IP is the server's current IP address,
1220 and dirport is its current directory port. XXXXorport
1222 "contact" SP string NL
1226 An arbitrary string describing how to contact the directory
1227 server's administrator. Administrators should include at least an
1228 email address and a PGP fingerprint.
1230 "legacy-key" SP FINGERPRINT NL
1234 Lists a fingerprint for an obsolete _identity_ key still used
1235 by this authority to keep older clients working. This option
1236 is used to keep key around for a little while in case the
1237 authorities need to migrate many identity keys at once.
1238 (Generally, this would only happen because of a security
1239 vulnerability that affected multiple authorities, like the
1240 Debian OpenSSL RNG bug of May 2008.)
1242 The authority section of a consensus contains groups the following items,
1243 in the order given, with one group for each authority that contributed to
1244 the consensus, with groups sorted by authority identity digest:
1246 "dir-source" SP nickname SP identity SP address SP IP SP dirport SP
1249 [Exactly once, at start]
1251 As in the authority section of a vote.
1253 "contact" SP string NL
1257 As in the authority section of a vote.
1259 "vote-digest" SP digest NL
1263 A digest of the vote from the authority that contributed to this
1264 consensus, as signed (that is, not including the signature).
1267 Each router status entry contains the following items. Router status
1268 entries are sorted in ascending order by identity digest.
1270 "r" SP nickname SP identity SP digest SP publication SP IP SP ORPort
1273 [At start, exactly once.]
1275 "Nickname" is the OR's nickname. "Identity" is a hash of its
1276 identity key, encoded in base64, with trailing equals sign(s)
1277 removed. "Digest" is a hash of its most recent descriptor as
1278 signed (that is, not including the signature), encoded in base64.
1279 "Publication" is the
1280 publication time of its most recent descriptor, in the form
1281 YYYY-MM-DD HH:MM:SS, in GMT. "IP" is its current IP address;
1282 ORPort is its current OR port, "DirPort" is it's current directory
1283 port, or "0" for "none".
1289 A series of space-separated status flags, in lexical order (as ASCII
1290 byte strings). Currently documented flags are:
1292 "Authority" if the router is a directory authority.
1293 "BadExit" if the router is believed to be useless as an exit node
1294 (because its ISP censors it, because it is behind a restrictive
1295 proxy, or for some similar reason).
1296 "BadDirectory" if the router is believed to be useless as a
1297 directory cache (because its directory port isn't working,
1298 its bandwidth is always throttled, or for some similar
1300 "Exit" if the router is more useful for building
1301 general-purpose exit circuits than for relay circuits. The
1302 path building algorithm uses this flag; see path-spec.txt.
1303 "Fast" if the router is suitable for high-bandwidth circuits.
1304 "Guard" if the router is suitable for use as an entry guard.
1305 "HSDir" if the router is considered a v2 hidden service directory.
1306 "Named" if the router's identity-nickname mapping is canonical,
1307 and this authority binds names.
1308 "Stable" if the router is suitable for long-lived circuits.
1309 "Running" if the router is currently usable.
1310 "Unnamed" if another router has bound the name used by this
1311 router, and this authority binds names.
1312 "Valid" if the router has been 'validated'.
1313 "V2Dir" if the router implements the v2 directory protocol.
1314 "V3Dir" if the router implements this protocol.
1320 The version of the Tor protocol that this server is running. If
1321 the value begins with "Tor" SP, the rest of the string is a Tor
1322 version number, and the protocol is "The Tor protocol as supported
1323 by the given version of Tor." Otherwise, if the value begins with
1324 some other string, Tor has upgraded to a more sophisticated
1325 protocol versioning system, and the protocol is "a version of the
1326 Tor protocol more recent than any we recognize."
1328 Directory authorities SHOULD omit version strings they receive from
1329 descriptors if they would cause "v" lines to be over 128 characters
1332 "w" SP "Bandwidth=" INT [SP "Measured=" INT] NL
1336 An estimate of the bandwidth of this server, in an arbitrary
1337 unit (currently kilobytes per second). Used to weight router
1340 Additionally, the Measured= keyword is present in votes by
1341 participating bandwidth measurement authorities to indicate
1342 a measured bandwidth currently produced by measuring stream
1345 Other weighting keywords may be added later.
1346 Clients MUST ignore keywords they do not recognize.
1348 "p" SP ("accept" / "reject") SP PortList NL
1352 PortList = PortOrRange
1353 PortList = PortList "," PortOrRange
1354 PortOrRange = INT "-" INT / INT
1356 A list of those ports that this router supports (if 'accept')
1357 or does not support (if 'reject') for exit to "most
1360 The footer section is delineated in all votes and consensuses supporting
1361 consensus method 9 and above with the following:
1363 "directory-footer" NL
1365 It contains two subsections, a bandwidths-weights line and a
1366 directory-signature.
1368 The bandwidths-weights line appears At Most Once for a consensus. It does
1369 not appear in votes.
1371 "bandwidth-weights" SP
1372 "Wbd=" INT SP "Wbe=" INT SP "Wbg=" INT SP "Wbm=" INT SP
1374 "Web=" INT SP "Wed=" INT SP "Wee=" INT SP "Weg=" INT SP "Wem=" INT SP
1375 "Wgb=" INT SP "Wgd=" INT SP "Wgg=" INT SP "Wgm=" INT SP
1376 "Wmb=" INT SP "Wmd=" INT SP "Wme=" INT SP "Wmg=" INT SP "Wmm=" INT NL
1378 These values represent the weights to apply to router bandwidths
1379 during path selection. They are sorted in lexical order (as ASCII byte
1380 strings). The integer values are divided by BW_WEIGHT_SCALE=10000 or
1381 the consensus param "bwweightscale". They are:
1383 Wgg - Weight for Guard-flagged nodes in the guard position
1384 Wgm - Weight for non-flagged nodes in the guard Position
1385 Wgd - Weight for Guard+Exit-flagged nodes in the guard Position
1387 Wmg - Weight for Guard-flagged nodes in the middle Position
1388 Wmm - Weight for non-flagged nodes in the middle Position
1389 Wme - Weight for Exit-flagged nodes in the middle Position
1390 Wmd - Weight for Guard+Exit flagged nodes in the middle Position
1392 Weg - Weight for Guard flagged nodes in the exit Position
1393 Wem - Weight for non-flagged nodes in the exit Position
1394 Wee - Weight for Exit-flagged nodes in the exit Position
1395 Wed - Weight for Guard+Exit-flagged nodes in the exit Position
1397 Wgb - Weight for BEGIN_DIR-supporting Guard-flagged nodes
1398 Wmb - Weight for BEGIN_DIR-supporting non-flagged nodes
1399 Web - Weight for BEGIN_DIR-supporting Exit-flagged nodes
1400 Wdb - Weight for BEGIN_DIR-supporting Guard+Exit-flagged nodes
1402 Wbg - Weight for Guard flagged nodes for BEGIN_DIR requests
1403 Wbm - Weight for non-flagged nodes for BEGIN_DIR requests
1404 Wbe - Weight for Exit-flagged nodes for BEGIN_DIR requests
1405 Wbd - Weight for Guard+Exit-flagged nodes for BEGIN_DIR requests
1407 These values are calculated as specified in Section 3.4.3.
1409 The signature contains the following item, which appears Exactly Once
1410 for a vote, and At Least Once for a consensus.
1412 "directory-signature" SP identity SP signing-key-digest NL Signature
1414 This is a signature of the status document, with the initial item
1415 "network-status-version", and the signature item
1416 "directory-signature", using the signing key. (In this case, we take
1417 the hash through the _space_ after directory-signature, not the
1418 newline: this ensures that all authorities sign the same thing.)
1419 "identity" is the hex-encoded digest of the authority identity key of
1420 the signing authority, and "signing-key-digest" is the hex-encoded
1421 digest of the current authority signing key of the signing authority.
1423 3.3. Assigning flags in a vote
1425 (This section describes how directory authorities choose which status
1426 flags to apply to routers, as of Tor 0.2.0.0-alpha-dev. Later directory
1427 authorities MAY do things differently, so long as clients keep working
1428 well. Clients MUST NOT depend on the exact behaviors in this section.)
1430 In the below definitions, a router is considered "active" if it is
1431 running, valid, and not hibernating.
1433 "Valid" -- a router is 'Valid' if it is running a version of Tor not
1434 known to be broken, and the directory authority has not blacklisted
1437 "Named" -- Directory authority administrators may decide to support name
1438 binding. If they do, then they must maintain a file of
1439 nickname-to-identity-key mappings, and try to keep this file consistent
1440 with other directory authorities. If they don't, they act as clients, and
1441 report bindings made by other directory authorities (name X is bound to
1442 identity Y if at least one binding directory lists it, and no directory
1443 binds X to some other Y'.) A router is called 'Named' if the router
1444 believes the given name should be bound to the given key.
1446 Two strategies exist on the current network for deciding on
1447 values for the Named flag. In the original version, server
1448 operators were asked to send nickname-identity pairs to a
1449 mailing list of Naming directory authorities operators. The
1450 operators were then supposed to add the pairs to their
1451 mapping files; in practice, they didn't get to this often.
1453 Newer Naming authorities run a script that registers routers
1454 in their mapping files once the routers have been online at
1455 least two weeks, no other router has that nickname, and no
1456 other router has wanted the nickname for a month. If a router
1457 has not been online for six months, the router is removed.
1459 "Unnamed" -- Directory authorities that support naming should vote for a
1460 router to be 'Unnamed' if its given nickname is mapped to a different
1463 "Running" -- A router is 'Running' if the authority managed to connect to
1464 it successfully within the last 30 minutes.
1466 "Stable" -- A router is 'Stable' if it is active, and either its Weighted
1467 MTBF is at least the median for known active routers or its Weighted MTBF
1468 corresponds to at least 7 days. Routers are never called Stable if they are
1469 running a version of Tor known to drop circuits stupidly. (0.1.1.10-alpha
1470 through 0.1.1.16-rc are stupid this way.)
1472 To calculate weighted MTBF, compute the weighted mean of the lengths
1473 of all intervals when the router was observed to be up, weighting
1474 intervals by $\alpha^n$, where $n$ is the amount of time that has
1475 passed since the interval ended, and $\alpha$ is chosen so that
1476 measurements over approximately one month old no longer influence the
1479 [XXXX what happens when we have less than 4 days of MTBF info.]
1481 "Exit" -- A router is called an 'Exit' iff it allows exits to at
1482 least two of the ports 80, 443, and 6667 and allows exits to at
1483 least one /8 address space.
1485 "Fast" -- A router is 'Fast' if it is active, and its bandwidth is
1486 either in the top 7/8ths for known active routers or at least 20KB/s.
1488 "Guard" -- A router is a possible 'Guard' if its Weighted Fractional
1489 Uptime is at least the median for "familiar" active routers, and if
1490 its bandwidth is at least median or at least 250KB/s.
1492 To calculate weighted fractional uptime, compute the fraction
1493 of time that the router is up in any given day, weighting so that
1494 downtime and uptime in the past counts less.
1496 A node is 'familiar' if 1/8 of all active nodes have appeared more
1497 recently than it, OR it has been around for a few weeks.
1499 "Authority" -- A router is called an 'Authority' if the authority
1500 generating the network-status document believes it is an authority.
1502 "V2Dir" -- A router supports the v2 directory protocol if it has an open
1503 directory port, and it is running a version of the directory protocol that
1504 supports the functionality clients need. (Currently, this is
1505 0.1.1.9-alpha or later.)
1507 "V3Dir" -- A router supports the v3 directory protocol if it has an open
1508 directory port, and it is running a version of the directory protocol that
1509 supports the functionality clients need. (Currently, this is
1510 0.2.0.?????-alpha or later.)
1512 "HSDir" -- A router is a v2 hidden service directory if it stores and
1513 serves v2 hidden service descriptors and the authority managed to connect
1514 to it successfully within the last 24 hours.
1516 Directory server administrators may label some servers or IPs as
1517 blacklisted, and elect not to include them in their network-status lists.
1519 Authorities SHOULD 'disable' any servers in excess of 3 on any single IP.
1520 When there are more than 3 to choose from, authorities should first prefer
1521 authorities to non-authorities, then prefer Running to non-Running, and
1522 then prefer high-bandwidth to low-bandwidth. To 'disable' a server, the
1523 authority *should* advertise it without the Running or Valid flag.
1525 Thus, the network-status vote includes all non-blacklisted,
1526 non-expired, non-superseded descriptors.
1528 The bandwidth in a "w" line should be taken as the best estimate
1529 of the router's actual capacity that the authority has. For now,
1530 this should be the lesser of the observed bandwidth and bandwidth
1531 rate limit from the router descriptor. It is given in kilobytes
1532 per second, and capped at some arbitrary value (currently 10 MB/s).
1534 The Measured= keyword on a "w" line vote is currently computed
1535 by multiplying the previous published consensus bandwidth by the
1536 ratio of the measured average node stream capacity to the network
1537 average. If 3 or more authorities provide a Measured= keyword for
1538 a router, the authorities produce a consensus containing a "w"
1539 Bandwidth= keyword equal to the median of the Measured= votes.
1541 The ports listed in a "p" line should be taken as those ports for
1542 which the router's exit policy permits 'most' addresses, ignoring any
1543 accept not for all addresses, ignoring all rejects for private
1544 netblocks. "Most" addresses are permitted if no more than 2^25
1545 IPv4 addresses (two /8 networks) were blocked. The list is encoded
1546 as described in 3.4.2.
1548 3.4. Computing a consensus from a set of votes
1550 Given a set of votes, authorities compute the contents of the consensus
1551 document as follows:
1553 The "valid-after", "valid-until", and "fresh-until" times are taken as
1554 the median of the respective values from all the votes.
1556 The times in the "voting-delay" line are taken as the median of the
1557 VoteSeconds and DistSeconds times in the votes.
1559 Known-flags is the union of all flags known by any voter.
1561 Entries are given on the "params" line for every keyword on which any
1562 authority voted. The values given are the low-median of all votes on
1565 "client-versions" and "server-versions" are sorted in ascending
1566 order; A version is recommended in the consensus if it is recommended
1567 by more than half of the voting authorities that included a
1568 client-versions or server-versions lines in their votes.
1570 The authority item groups (dir-source, contact, fingerprint,
1571 vote-digest) are taken from the votes of the voting
1572 authorities. These groups are sorted by the digests of the
1573 authorities identity keys, in ascending order. If the consensus
1574 method is 3 or later, a dir-source line must be included for
1575 every vote with legacy-key entry, using the legacy-key's
1576 fingerprint, the voter's ordinary nickname with the string
1577 "-legacy" appended, and all other fields as from the original
1578 vote's dir-source line.
1580 A router status entry:
1581 * is included in the result if some router status entry with the same
1582 identity is included by more than half of the authorities (total
1583 authorities, not just those whose votes we have).
1585 * For any given identity, we include at most one router status entry.
1587 * A router entry has a flag set if that is included by more than half
1588 of the authorities who care about that flag.
1590 * Two router entries are "the same" if they have the same
1591 <descriptor digest, published time, nickname, IP, ports> tuple.
1592 We choose the tuple for a given router as whichever tuple appears
1593 for that router in the most votes. We break ties first in favor of
1594 the more recently published, then in favor of smaller server
1597 * The Named flag appears if it is included for this routerstatus by
1598 _any_ authority, and if all authorities that list it list the same
1599 nickname. However, if consensus-method 2 or later is in use, and
1600 any authority calls this identity/nickname pair Unnamed, then
1601 this routerstatus does not get the Named flag.
1603 * If consensus-method 2 or later is in use, the Unnamed flag is
1604 set for a routerstatus if any authorities have voted for a different
1605 identities to be Named with that nickname, or if any authority
1606 lists that nickname/ID pair as Unnamed.
1608 (With consensus-method 1, Unnamed is set like any other flag.)
1610 * The version is given as whichever version is listed by the most
1611 voters, with ties decided in favor of more recent versions.
1613 * If consensus-method 4 or later is in use, then routers that
1614 do not have the Running flag are not listed at all.
1616 * If consensus-method 5 or later is in use, then the "w" line
1617 is generated using a low-median of the bandwidth values from
1618 the votes that included "w" lines for this router.
1620 * If consensus-method 5 or later is in use, then the "p" line
1621 is taken from the votes that have the same policy summary
1622 for the descriptor we are listing. (They should all be the
1623 same. If they are not, we pick the most commonly listed
1624 one, breaking ties in favor of the lexicographically larger
1625 vote.) The port list is encoded as specified in 3.4.2.
1627 * If consensus-method 6 or later is in use and if 3 or more
1628 authorities provide a Measured= keyword in their votes for
1629 a router, the authorities produce a consensus containing a
1630 Bandwidth= keyword equal to the median of the Measured= votes.
1632 * If consensus-method 7 or later is in use, the params line is
1633 included in the output.
1635 * If the consensus method is under 11, bad exits are considered as
1636 possible exits when computing bandwidth weights. Otherwise, if
1637 method 11 or later is in use, any router that is determined to get
1638 the BadExit flag doesn't count when we're calculating weights.
1640 The signatures at the end of a consensus document are sorted in
1641 ascending order by identity digest.
1643 All ties in computing medians are broken in favor of the smaller or
1646 3.4.1. Forward compatibility
1648 Future versions of Tor will need to include new information in the
1649 consensus documents, but it is important that all authorities (or at least
1650 half) generate and sign the same signed consensus.
1652 To achieve this, authorities list in their votes their supported methods
1653 for generating consensuses from votes. Later methods will be assigned
1654 higher numbers. Currently recognized methods:
1655 "1" -- The first implemented version.
1656 "2" -- Added support for the Unnamed flag.
1657 "3" -- Added legacy ID key support to aid in authority ID key rollovers
1658 "4" -- No longer list routers that are not running in the consensus
1659 "5" -- adds support for "w" and "p" lines.
1660 "6" -- Prefers measured bandwidth values rather than advertised
1661 "7" -- Provides keyword=integer pairs of consensus parameters
1662 "8" -- Provides microdescriptor summaries
1663 "9" -- Provides weights for selecting flagged routers in paths
1664 "10" -- Fixes edge case bugs in router flag selection weights
1665 "11" -- Don't consider BadExits when calculating bandwidth weights
1668 Before generating a consensus, an authority must decide which consensus
1669 method to use. To do this, it looks for the highest version number
1670 supported by more than 2/3 of the authorities voting. If it supports this
1671 method, then it uses it. Otherwise, it falls back to method 1.
1673 (The consensuses generated by new methods must be parsable by
1674 implementations that only understand the old methods, and must not cause
1675 those implementations to compromise their anonymity. This is a means for
1676 making changes in the contents of consensus; not for making
1677 backward-incompatible changes in their format.)
1679 3.4.2. Encoding port lists
1681 Whether the summary shows the list of accepted ports or the list of
1682 rejected ports depends on which list is shorter (has a shorter string
1683 representation). In case of ties we choose the list of accepted
1684 ports. As an exception to this rule an allow-all policy is
1685 represented as "accept 1-65535" instead of "reject " and a reject-all
1686 policy is similarly given as "reject 1-65535".
1688 Summary items are compressed, that is instead of "80-88,89-100" there
1689 only is a single item of "80-100", similarly instead of "20,21" a
1690 summary will say "20-21".
1692 Port lists are sorted in ascending order.
1694 The maximum allowed length of a policy summary (including the "accept "
1695 or "reject ") is 1000 characters. If a summary exceeds that length we
1696 use an accept-style summary and list as much of the port list as is
1697 possible within these 1000 bytes. [XXXX be more specific.]
1699 3.4.3. Computing Bandwidth Weights
1701 Let weight_scale = 10000
1703 Let G be the total bandwidth for Guard-flagged nodes.
1704 Let M be the total bandwidth for non-flagged nodes.
1705 Let E be the total bandwidth for Exit-flagged nodes.
1706 Let D be the total bandwidth for Guard+Exit-flagged nodes.
1709 Let Wgd be the weight for choosing a Guard+Exit for the guard position.
1710 Let Wmd be the weight for choosing a Guard+Exit for the middle position.
1711 Let Wed be the weight for choosing a Guard+Exit for the exit position.
1713 Let Wme be the weight for choosing an Exit for the middle position.
1714 Let Wmg be the weight for choosing a Guard for the middle position.
1716 Let Wgg be the weight for choosing a Guard for the guard position.
1717 Let Wee be the weight for choosing an Exit for the exit position.
1719 Balanced network conditions then arise from solutions to the following
1720 system of equations:
1722 Wgg*G + Wgd*D == M + Wmd*D + Wme*E + Wmg*G (guard bw = middle bw)
1723 Wgg*G + Wgd*D == Wee*E + Wed*D (guard bw = exit bw)
1724 Wed*D + Wmd*D + Wgd*D == D (aka: Wed+Wmd+Wdg = 1)
1725 Wmg*G + Wgg*G == G (aka: Wgg = 1-Wmg)
1726 Wme*E + Wee*E == E (aka: Wee = 1-Wme)
1728 We are short 2 constraints with the above set. The remaining constraints
1729 come from examining different cases of network load. The following
1730 constraints are used in consensus method 10 and above. There are another
1731 incorrect and obsolete set of constraints used for these same cases in
1732 consensus method 9. For those, see dir-spec.txt in Tor 0.2.2.10-alpha
1735 Case 1: E >= T/3 && G >= T/3 (Neither Exit nor Guard Scarce)
1737 In this case, the additional two constraints are: Wmg == Wmd,
1740 This leads to the solution:
1741 Wgd = weight_scale/3
1742 Wed = weight_scale/3
1743 Wmd = weight_scale/3
1744 Wee = (weight_scale*(E+G+M))/(3*E)
1745 Wme = weight_scale - Wee
1746 Wmg = (weight_scale*(2*G-E-M))/(3*G)
1747 Wgg = weight_scale - Wmg
1749 Case 2: E < T/3 && G < T/3 (Both are scarce)
1751 Let R denote the more scarce class (Rare) between Guard vs Exit.
1752 Let S denote the less scarce class.
1756 In this subcase, we simply devote all of D bandwidth to the
1759 Wgg = Wee = weight_scale
1760 Wmg = Wme = Wmd = 0;
1770 In this case, if M <= T/3, we have enough bandwidth to try to achieve
1771 a balancing condition.
1773 Add constraints Wgg = 1, Wmd == Wgd to maximize bandwidth in the guard
1774 position while still allowing exits to be used as middle nodes:
1776 Wee = (weight_scale*(E - G + M))/E
1777 Wed = (weight_scale*(D - 2*E + 4*G - 2*M))/(3*D)
1778 Wme = (weight_scale*(G-M))/E
1781 Wmd = (weight_scale - Wed)/2
1782 Wgd = (weight_scale - Wed)/2
1784 If this system ends up with any values out of range (ie negative, or
1785 above weight_scale), use the constraints Wgg == 1 and Wee == 1, since
1786 both those positions are scarce:
1790 Wed = (weight_scale*(D - 2*E + G + M))/(3*D)
1791 Wmd = (weight_Scale*(D - 2*M + G + E))/(3*D)
1794 Wgd = weight_scale - Wed - Wmd
1796 If M > T/3, then the Wmd weight above will become negative. Set it to 0
1799 Wgd = weight_scale - Wed
1801 Case 3: One of E < T/3 or G < T/3
1803 Let S be the scarce class (of E or G).
1805 Subcase a: (S+D) < T/3:
1807 Wgg = Wgd = weight_scale;
1808 Wmd = Wed = Wmg = 0;
1809 // Minor subcase, if E is more scarce than M,
1810 // keep its bandwidth in place.
1812 else Wme = (weight_scale*(E-M))/(2*E);
1813 Wee = weight_scale-Wme;
1815 Wee = Wed = weight_scale;
1816 Wmd = Wgd = Wme = 0;
1817 // Minor subcase, if G is more scarce than M,
1818 // keep its bandwidth in place.
1820 else Wmg = (weight_scale*(G-M))/(2*G);
1821 Wgg = weight_scale-Wmg;
1823 Subcase b: (S+D) >= T/3
1825 Add constraints Wgg = 1, Wmd == Wed to maximize bandwidth
1826 in the guard position, while still allowing exits to be
1827 used as middle nodes:
1829 Wgd = (weight_scale*(D - 2*G + E + M))/(3*D)
1831 Wee = (weight_scale*(E+M))/(2*E)
1832 Wme = weight_scale - Wee
1833 Wmd = (weight_scale - Wgd)/2
1834 Wed = (weight_scale - Wgd)/2
1836 Add constraints Wee == 1, Wmd == Wgd to maximize bandwidth
1837 in the exit position:
1839 Wed = (weight_scale*(D - 2*E + G + M))/(3*D);
1841 Wgg = (weight_scale*(G+M))/(2*G);
1842 Wmg = weight_scale - Wgg;
1843 Wmd = (weight_scale - Wed)/2;
1844 Wgd = (weight_scale - Wed)/2;
1846 To ensure consensus, all calculations are performed using integer math
1847 with a fixed precision determined by the bwweightscale consensus
1848 parameter (defaults at 10000, Min: 1, Max: INT32_MAX).
1850 For future balancing improvements, Tor clients support 11 additional weights
1851 for directory requests and middle weighting. These weights are currently
1852 set at weight_scale, with the exception of the following groups of
1855 Directory requests use middle weights:
1856 Wbd=Wmd, Wbg=Wmg, Wbe=Wme, Wbm=Wmm
1858 Handle bridges and strange exit policies:
1859 Wgm=Wgg, Wem=Wee, Weg=Wed
1861 3.5. Detached signatures
1863 Assuming full connectivity, every authority should compute and sign the
1864 same consensus directory in each period. Therefore, it isn't necessary to
1865 download the consensus computed by each authority; instead, the
1866 authorities only push/fetch each others' signatures. A "detached
1867 signature" document contains items as follows:
1869 "consensus-digest" SP Digest NL
1871 [At start, at most once.]
1873 The digest of the consensus being signed.
1875 "valid-after" SP YYYY-MM-DD SP HH:MM:SS NL
1876 "fresh-until" SP YYYY-MM-DD SP HH:MM:SS NL
1877 "valid-until" SP YYYY-MM-DD SP HH:MM:SS NL
1879 [As in the consensus]
1881 "directory-signature"
1883 [As in the consensus; the signature object is the same as in the
1884 consensus document.]
1887 4. Directory server operation
1889 All directory authorities and directory caches ("directory servers")
1890 implement this section, except as noted.
1892 4.1. Accepting uploads (authorities only)
1894 When a router posts a signed descriptor to a directory authority, the
1895 authority first checks whether it is well-formed and correctly
1896 self-signed. If it is, the authority next verifies that the nickname
1897 in question is not already assigned to a router with a different
1899 Finally, the authority MAY check that the router is not blacklisted
1900 because of its key, IP, or another reason.
1902 If the descriptor passes these tests, and the authority does not already
1903 have a descriptor for a router with this public key, it accepts the
1904 descriptor and remembers it.
1906 If the authority _does_ have a descriptor with the same public key, the
1907 newly uploaded descriptor is remembered if its publication time is more
1908 recent than the most recent old descriptor for that router, and either:
1909 - There are non-cosmetic differences between the old descriptor and the
1911 - Enough time has passed between the descriptors' publication times.
1912 (Currently, 12 hours.)
1914 Differences between router descriptors are "non-cosmetic" if they would be
1915 sufficient to force an upload as described in section 2 above.
1917 Note that the "cosmetic difference" test only applies to uploaded
1918 descriptors, not to descriptors that the authority downloads from other
1921 When a router posts a signed extra-info document to a directory authority,
1922 the authority again checks it for well-formedness and correct signature,
1923 and checks that its matches the extra-info-digest in some router
1924 descriptor that it believes is currently useful. If so, it accepts it and
1925 stores it and serves it as requested. If not, it drops it.
1927 4.2. Voting (authorities only)
1929 Authorities divide time into Intervals. Authority administrators SHOULD
1930 try to all pick the same interval length, and SHOULD pick intervals that
1931 are commonly used divisions of time (e.g., 5 minutes, 15 minutes, 30
1932 minutes, 60 minutes, 90 minutes). Voting intervals SHOULD be chosen to
1933 divide evenly into a 24-hour day.
1935 Authorities SHOULD act according to interval and delays in the
1936 latest consensus. Lacking a latest consensus, they SHOULD default to a
1937 30-minute Interval, a 5 minute VotingDelay, and a 5 minute DistDelay.
1939 Authorities MUST take pains to ensure that their clocks remain accurate
1940 within a few seconds. (Running NTP is usually sufficient.)
1942 The first voting period of each day begins at 00:00 (midnight) GMT. If
1943 the last period of the day would be truncated by one-half or more, it is
1944 merged with the second-to-last period.
1946 An authority SHOULD publish its vote immediately at the start of each voting
1947 period (minus VoteSeconds+DistSeconds). It does this by making it
1949 http://<hostname>/tor/status-vote/next/authority.z
1950 and sending it in an HTTP POST request to each other authority at the URL
1951 http://<hostname>/tor/post/vote
1953 If, at the start of the voting period, minus DistSeconds, an authority
1954 does not have a current statement from another authority, the first
1955 authority downloads the other's statement.
1957 Once an authority has a vote from another authority, it makes it available
1959 http://<hostname>/tor/status-vote/next/<fp>.z
1960 where <fp> is the fingerprint of the other authority's identity key.
1962 http://<hostname>/tor/status-vote/next/d/<d>.z
1963 where <d> is the digest of the vote document.
1965 The consensus status, along with as many signatures as the server
1966 currently knows, should be available at
1967 http://<hostname>/tor/status-vote/next/consensus.z
1968 All of the detached signatures it knows for consensus status should be
1970 http://<hostname>/tor/status-vote/next/consensus-signatures.z
1972 Once there are enough signatures, or once the voting period starts,
1973 these documents are available at
1974 http://<hostname>/tor/status-vote/current/consensus.z
1976 http://<hostname>/tor/status-vote/current/consensus-signatures.z
1977 [XXX current/consensus-signatures is not currently implemented, as it
1978 is not used in the voting protocol.]
1980 The other vote documents are analogously made available under
1981 http://<hostname>/tor/status-vote/current/authority.z
1982 http://<hostname>/tor/status-vote/current/<fp>.z
1983 http://<hostname>/tor/status-vote/current/d/<d>.z
1984 once the consensus is complete.
1986 Once an authority has computed and signed a consensus network status, it
1987 should send its detached signature to each other authority in an HTTP POST
1989 http://<hostname>/tor/post/consensus-signature
1991 [XXX Note why we support push-and-then-pull.]
1993 [XXX possible future features include support for downloading old
1996 4.3. Downloading consensus status documents (caches only)
1998 All directory servers (authorities and caches) try to keep a recent
1999 network-status consensus document to serve to clients. A cache ALWAYS
2000 downloads a network-status consensus if any of the following are true:
2001 - The cache has no consensus document.
2002 - The cache's consensus document is no longer valid.
2003 Otherwise, the cache downloads a new consensus document at a randomly
2004 chosen time in the first half-interval after its current consensus
2005 stops being fresh. (This time is chosen at random to avoid swarming
2006 the authorities at the start of each period. The interval size is
2007 inferred from the difference between the valid-after time and the
2008 fresh-until time on the consensus.)
2010 [For example, if a cache has a consensus that became valid at 1:00,
2011 and is fresh until 2:00, that cache will fetch a new consensus at
2012 a random time between 2:00 and 2:30.]
2014 4.4. Downloading and storing router descriptors (authorities and caches)
2016 Periodically (currently, every 10 seconds), directory servers check
2017 whether there are any specific descriptors that they do not have and that
2018 they are not currently trying to download. Caches identify these
2019 descriptors by hash in the recent network-status consensus documents;
2020 authorities identify them by hash in vote (if publication date is more
2021 recent than the descriptor we currently have).
2023 [XXXX need a way to fetch descriptors ahead of the vote? v2 status docs can
2026 If so, the directory server launches requests to the authorities for these
2027 descriptors, such that each authority is only asked for descriptors listed
2028 in its most recent vote (if the requester is an authority) or in the
2029 consensus (if the requester is a cache). If we're an authority, and more
2030 than one authority lists the descriptor, we choose which to ask at random.
2032 If one of these downloads fails, we do not try to download that descriptor
2033 from the authority that failed to serve it again unless we receive a newer
2034 network-status (consensus or vote) from that authority that lists the same
2037 Directory servers must potentially cache multiple descriptors for each
2038 router. Servers must not discard any descriptor listed by any recent
2039 consensus. If there is enough space to store additional descriptors,
2040 servers SHOULD try to hold those which clients are likely to download the
2041 most. (Currently, this is judged based on the interval for which each
2042 descriptor seemed newest.)
2043 [XXXX define recent]
2045 Authorities SHOULD NOT download descriptors for routers that they would
2046 immediately reject for reasons listed in 3.1.
2048 4.5. Downloading and storing extra-info documents
2050 All authorities, and any cache that chooses to cache extra-info documents,
2051 and any client that uses extra-info documents, should implement this
2054 Note that generally, clients don't need extra-info documents.
2056 Periodically, the Tor instance checks whether it is missing any extra-info
2057 documents: in other words, if it has any router descriptors with an
2058 extra-info-digest field that does not match any of the extra-info
2059 documents currently held. If so, it downloads whatever extra-info
2060 documents are missing. Caches download from authorities; non-caches try
2061 to download from caches. We follow the same splitting and back-off rules
2062 as in 4.4 (if a cache) or 5.3 (if a client).
2064 4.6. General-use HTTP URLs
2066 "Fingerprints" in these URLs are base-16-encoded SHA1 hashes.
2068 The most recent v3 consensus should be available at:
2069 http://<hostname>/tor/status-vote/current/consensus.z
2071 Starting with Tor version 0.2.1.1-alpha is also available at:
2072 http://<hostname>/tor/status-vote/current/consensus/<F1>+<F2>+<F3>.z
2074 Where F1, F2, etc. are authority identity fingerprints the client trusts.
2075 Servers will only return a consensus if more than half of the requested
2076 authorities have signed the document, otherwise a 404 error will be sent
2077 back. The fingerprints can be shortened to a length of any multiple of
2078 two, using only the leftmost part of the encoded fingerprint. Tor uses
2079 3 bytes (6 hex characters) of the fingerprint.
2081 Clients SHOULD sort the fingerprints in ascending order. Server MUST
2084 Clients SHOULD use this format when requesting consensus documents from
2085 directory authority servers and from caches running a version of Tor
2086 that is known to support this URL format.
2088 A concatenated set of all the current key certificates should be available
2090 http://<hostname>/tor/keys/all.z
2092 The key certificate for this server (if it is an authority) should be
2094 http://<hostname>/tor/keys/authority.z
2096 The key certificate for an authority whose authority identity fingerprint
2097 is <F> should be available at:
2098 http://<hostname>/tor/keys/fp/<F>.z
2100 The key certificate whose signing key fingerprint is <F> should be
2102 http://<hostname>/tor/keys/sk/<F>.z
2104 The key certificate whose identity key fingerprint is <F> and whose signing
2105 key fingerprint is <S> should be available at:
2107 http://<hostname>/tor/keys/fp-sk/<F>-<S>.z
2109 (As usual, clients may request multiple certificates using:
2110 http://<hostname>/tor/keys/fp-sk/<F1>-<S1>+<F2>-<S2>.z )
2111 [The above fp-sk format was not supported before Tor 0.2.1.9-alpha.]
2113 The most recent descriptor for a server whose identity key has a
2114 fingerprint of <F> should be available at:
2115 http://<hostname>/tor/server/fp/<F>.z
2117 The most recent descriptors for servers with identity fingerprints
2118 <F1>,<F2>,<F3> should be available at:
2119 http://<hostname>/tor/server/fp/<F1>+<F2>+<F3>.z
2121 (NOTE: Implementations SHOULD NOT download descriptors by identity key
2122 fingerprint. This allows a corrupted server (in collusion with a cache) to
2123 provide a unique descriptor to a client, and thereby partition that client
2124 from the rest of the network.)
2126 The server descriptor with (descriptor) digest <D> (in hex) should be
2128 http://<hostname>/tor/server/d/<D>.z
2130 The most recent descriptors with digests <D1>,<D2>,<D3> should be
2132 http://<hostname>/tor/server/d/<D1>+<D2>+<D3>.z
2134 The most recent descriptor for this server should be at:
2135 http://<hostname>/tor/server/authority.z
2136 [Nothing in the Tor protocol uses this resource yet, but it is useful
2137 for debugging purposes. Also, the official Tor implementations
2138 (starting at 0.1.1.x) use this resource to test whether a server's
2139 own DirPort is reachable.]
2141 A concatenated set of the most recent descriptors for all known servers
2142 should be available at:
2143 http://<hostname>/tor/server/all.z
2145 Extra-info documents are available at the URLS
2146 http://<hostname>/tor/extra/d/...
2147 http://<hostname>/tor/extra/fp/...
2148 http://<hostname>/tor/extra/all[.z]
2149 http://<hostname>/tor/extra/authority[.z]
2150 (As for /tor/server/ URLs: supports fetching extra-info
2151 documents by their digest, by the fingerprint of their servers,
2152 or all at once. When serving by fingerprint, we serve the
2153 extra-info that corresponds to the descriptor we would serve by
2154 that fingerprint. Only directory authorities of version
2155 0.2.0.1-alpha or later are guaranteed to support the first
2156 three classes of URLs. Caches may support them, and MUST
2157 support them if they have advertised "caches-extra-info".)
2159 For debugging, directories SHOULD expose non-compressed objects at URLs like
2160 the above, but without the final ".z".
2161 Clients MUST handle compressed concatenated information in two forms:
2162 - A concatenated list of zlib-compressed objects.
2163 - A zlib-compressed concatenated list of objects.
2164 Directory servers MAY generate either format: the former requires less
2165 CPU, but the latter requires less bandwidth.
2167 Clients SHOULD use upper case letters (A-F) when base16-encoding
2168 fingerprints. Servers MUST accept both upper and lower case fingerprints
2171 5. Client operation: downloading information
2173 Every Tor that is not a directory server (that is, those that do
2174 not have a DirPort set) implements this section.
2176 5.1. Downloading network-status documents
2178 Each client maintains a list of directory authorities. Insofar as
2179 possible, clients SHOULD all use the same list.
2181 Clients try to have a live consensus network-status document at all times.
2182 A network-status document is "live" if the time in its valid-until field
2185 If a client is missing a live network-status document, it tries to fetch
2186 it from a directory cache (or from an authority if it knows no caches).
2187 On failure, the client waits briefly, then tries that network-status
2188 document again from another cache. The client does not build circuits
2189 until it has a live network-status consensus document, and it has
2190 descriptors for more than 1/4 of the routers that it believes are running.
2192 (Note: clients can and should pick caches based on the network-status
2193 information they have: once they have first fetched network-status info
2194 from an authority, they should not need to go to the authority directly
2197 To avoid swarming the caches whenever a consensus expires, the
2198 clients download new consensuses at a randomly chosen time after the
2199 caches are expected to have a fresh consensus, but before their
2200 consensus will expire. (This time is chosen uniformly at random from
2201 the interval between the time 3/4 into the first interval after the
2202 consensus is no longer fresh, and 7/8 of the time remaining after
2203 that before the consensus is invalid.)
2205 [For example, if a cache has a consensus that became valid at 1:00,
2206 and is fresh until 2:00, and expires at 4:00, that cache will fetch
2207 a new consensus at a random time between 2:45 and 3:50, since 3/4
2208 of the one-hour interval is 45 minutes, and 7/8 of the remaining 75
2209 minutes is 65 minutes.]
2211 5.2. Downloading and storing router descriptors
2213 Clients try to have the best descriptor for each router. A descriptor is
2215 * It is listed in the consensus network-status document.
2217 Periodically (currently every 10 seconds) clients check whether there are
2218 any "downloadable" descriptors. A descriptor is downloadable if:
2219 - It is the "best" descriptor for some router.
2220 - The descriptor was published at least 10 minutes in the past.
2221 (This prevents clients from trying to fetch descriptors that the
2222 mirrors have probably not yet retrieved and cached.)
2223 - The client does not currently have it.
2224 - The client is not currently trying to download it.
2225 - The client would not discard it immediately upon receiving it.
2226 - The client thinks it is running and valid (see 6.1 below).
2228 If at least 16 known routers have downloadable descriptors, or if
2229 enough time (currently 10 minutes) has passed since the last time the
2230 client tried to download descriptors, it launches requests for all
2231 downloadable descriptors, as described in 5.3 below.
2233 When a descriptor download fails, the client notes it, and does not
2234 consider the descriptor downloadable again until a certain amount of time
2235 has passed. (Currently 0 seconds for the first failure, 60 seconds for the
2236 second, 5 minutes for the third, 10 minutes for the fourth, and 1 day
2237 thereafter.) Periodically (currently once an hour) clients reset the
2240 Clients retain the most recent descriptor they have downloaded for each
2241 router so long as it is not too old (currently, 48 hours), OR so long as
2242 no better descriptor has been downloaded for the same router.
2244 [Versions of Tor before 0.1.2.3-alpha would discard descriptors simply for
2245 being published too far in the past.] [The code seems to discard
2246 descriptors in all cases after they're 5 days old. True? -RD]
2248 5.3. Managing downloads
2250 When a client has no consensus network-status document, it downloads it
2251 from a randomly chosen authority. In all other cases, the client
2252 downloads from caches randomly chosen from among those believed to be V2
2253 directory servers. (This information comes from the network-status
2254 documents; see 6 below.)
2256 When downloading multiple router descriptors, the client chooses multiple
2258 - At least 3 different mirrors are used, except when this would result
2259 in more than one request for under 4 descriptors.
2260 - No more than 128 descriptors are requested from a single mirror.
2261 - Otherwise, as few mirrors as possible are used.
2262 After choosing mirrors, the client divides the descriptors among them
2265 After receiving any response client MUST discard any network-status
2266 documents and descriptors that it did not request.
2268 6. Using directory information
2270 Everyone besides directory authorities uses the approaches in this section
2271 to decide which servers to use and what their keys are likely to be.
2272 (Directory authorities just believe their own opinions, as in 3.1 above.)
2274 6.1. Choosing routers for circuits.
2276 Circuits SHOULD NOT be built until the client has enough directory
2277 information: a live consensus network status [XXXX fallback?] and
2278 descriptors for at least 1/4 of the servers believed to be running.
2280 A server is "listed" if it is included by the consensus network-status
2281 document. Clients SHOULD NOT use unlisted servers.
2283 These flags are used as follows:
2285 - Clients SHOULD NOT use non-'Valid' or non-'Running' routers unless
2288 - Clients SHOULD NOT use non-'Fast' routers for any purpose other than
2289 very-low-bandwidth circuits (such as introduction circuits).
2291 - Clients SHOULD NOT use non-'Stable' routers for circuits that are
2292 likely to need to be open for a very long time (such as those used for
2293 IRC or SSH connections).
2295 - Clients SHOULD NOT choose non-'Guard' nodes when picking entry guard
2298 - Clients SHOULD NOT download directory information from non-'V2Dir'
2301 See the "path-spec.txt" document for more details.
2303 6.2. Managing naming
2305 In order to provide human-memorable names for individual server
2306 identities, some directory servers bind names to IDs. Clients handle
2309 When a client encounters a name it has not mapped before:
2311 If the consensus lists any router with that name as "Named", or if
2312 consensus-method 2 or later is in use and the consensus lists any
2313 router with that name as having the "Unnamed" flag, then the name is
2314 bound. (It's bound to the ID listed in the entry with the Named,
2315 or to an unknown ID if no name is found.)
2317 When the user refers to a bound name, the implementation SHOULD provide
2318 only the router with ID bound to that name, and no other router, even
2319 if the router with the right ID can't be found.
2321 When a user tries to refer to a non-bound name, the implementation SHOULD
2322 warn the user. After warning the user, the implementation MAY use any
2323 router that advertises the name.
2325 Not every router needs a nickname. When a router doesn't configure a
2326 nickname, it publishes with the default nickname "Unnamed". Authorities
2327 SHOULD NOT ever mark a router with this nickname as Named; client software
2328 SHOULD NOT ever use a router in response to a user request for a router
2331 6.3. Software versions
2333 An implementation of Tor SHOULD warn when it has fetched a consensus
2334 network-status, and it is running a software version not listed.
2336 6.4. Warning about a router's status.
2338 If a router tries to publish its descriptor to a Naming authority
2339 that has its nickname mapped to another key, the router SHOULD
2340 warn the operator that it is either using the wrong key or is using
2341 an already claimed nickname.
2343 If a router has fetched a consensus document,, and the
2344 authorities do not publish a binding for the router's nickname, the
2345 router MAY remind the operator that the chosen nickname is not
2346 bound to this key at the authorities, and suggest contacting the
2347 authority operators.
2351 6.5. Router protocol versions
2353 A client should believe that a router supports a given feature if that
2354 feature is supported by the router or protocol versions in more than half
2355 of the live networkstatuses' "v" entries for that router. In other words,
2356 if the "v" entries for some router are:
2357 v Tor 0.0.8pre1 (from authority 1)
2358 v Tor 0.1.2.11 (from authority 2)
2359 v FutureProtocolDescription 99 (from authority 3)
2360 then the client should believe that the router supports any feature
2361 supported by 0.1.2.11.
2363 This is currently equivalent to believing the median declared version for
2364 a router in all live networkstatuses.
2366 7. Standards compliance
2368 All clients and servers MUST support HTTP 1.0. Clients and servers MAY
2369 support later versions of HTTP as well.
2373 Servers MAY set the Content-Length: header. Servers SHOULD set
2374 Content-Encoding to "deflate" or "identity".
2376 Servers MAY include an X-Your-Address-Is: header, whose value is the
2377 apparent IP address of the client connecting to them (as a dotted quad).
2378 For directory connections tunneled over a BEGIN_DIR stream, servers SHOULD
2379 report the IP from which the circuit carrying the BEGIN_DIR stream reached
2380 them. [Servers before version 0.1.2.5-alpha reported 127.0.0.1 for all
2381 BEGIN_DIR-tunneled connections.]
2383 Servers SHOULD disable caching of multiple network statuses or multiple
2384 router descriptors. Servers MAY enable caching of single descriptors,
2385 single network statuses, the list of all router descriptors, a v1
2386 directory, or a v1 running routers document. XXX mention times.
2388 7.2. HTTP status codes
2390 Tor delivers the following status codes. Some were chosen without much
2391 thought; other code SHOULD NOT rely on specific status codes yet.
2393 200 -- the operation completed successfully
2394 -- the user requested statuses or serverdescs, and none of the ones we
2395 requested were found (0.2.0.4-alpha and earlier).
2397 304 -- the client specified an if-modified-since time, and none of the
2398 requested resources have changed since that time.
2400 400 -- the request is malformed, or
2401 -- the URL is for a malformed variation of one of the URLs we support,
2403 -- the client tried to post to a non-authority, or
2404 -- the authority rejected a malformed posted document, or
2406 404 -- the requested document was not found.
2407 -- the user requested statuses or serverdescs, and none of the ones
2408 requested were found (0.2.0.5-alpha and later).
2410 503 -- we are declining the request in order to save bandwidth
2411 -- user requested some items that we ordinarily generate or store,
2412 but we do not have any available.
2414 9. Backward compatibility and migration plans
2416 Until Tor versions before 0.1.1.x are completely obsolete, directory
2417 authorities should generate, and mirrors should download and cache, v1
2418 directories and running-routers lists, and allow old clients to download
2419 them. These documents and the rules for retrieving, serving, and caching
2420 them are described in dir-spec-v1.txt.
2422 Until Tor versions before 0.2.0.x are completely obsolete, directory
2423 authorities should generate, mirrors should download and cache, v2
2424 network-status documents, and allow old clients to download them.
2425 Additionally, all directory servers and caches should download, store, and
2426 serve any router descriptor that is required because of v2 network-status
2427 documents. These documents and the rules for retrieving, serving, and
2428 caching them are described in dir-spec-v1.txt.
2430 A. Consensus-negotiation timeline.
2432 Period begins: this is the Published time.
2433 Everybody sends votes
2434 Reconciliation: everybody tries to fetch missing votes.
2435 consensus may exist at this point.
2436 End of voting period:
2437 everyone swaps signatures.
2438 Now it's okay for caches to download
2439 Now it's okay for clients to download.
2441 Valid-after/valid-until switchover