| blob | 0a2b01db681ef67112c13ca4d7bd2999fc1a5078 |
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 AS OF 14 JUNE 2007, THIS SPECIFICATION HAS NOT YET BEEN COMPLETELY
22 IMPLEMENTED, OR COMPLETELY COMPLETED.
24 XXX when to download certificates.
25 XXX timeline
26 XXX fill in XXXXs
28 0.1. History
30 The earliest versions of Onion Routing shipped with a list of known
31 routers and their keys. When the set of routers changed, users needed to
32 fetch a new list.
34 The Version 1 Directory protocol
35 --------------------------------
37 Early versions of Tor (0.0.2) introduced "Directory authorities": servers
38 that served signed "directory" documents containing a list of signed
39 "router descriptors", along with short summary of the status of each
40 router. Thus, clients could get up-to-date information on the state of
41 the network automatically, and be certain that the list they were getting
42 was attested by a trusted directory authority.
44 Later versions (0.0.8) added directory caches, which download
45 directories from the authorities and serve them to clients. Non-caches
46 fetch from the caches in preference to fetching from the authorities, thus
47 distributing bandwidth requirements.
49 Also added during the version 1 directory protocol were "router status"
50 documents: short documents that listed only the up/down status of the
51 routers on the network, rather than a complete list of all the
52 descriptors. Clients and caches would fetch these documents far more
53 frequently than they would fetch full directories.
55 The Version 2 Directory Protocol
56 --------------------------------
58 During the Tor 0.1.1.x series, Tor revised its handling of directory
59 documents in order to address two major problems:
61 * Directories had grown quite large (over 1MB), and most directory
62 downloads consisted mainly of router descriptors that clients
63 already had.
65 * Every directory authority was a trust bottleneck: if a single
66 directory authority lied, it could make clients believe for a time
67 an arbitrarily distorted view of the Tor network. (Clients
68 trusted the most recent signed document they downloaded.) Thus,
69 adding more authorities would make the system less secure, not
70 more.
72 To address these, we extended the directory protocol so that
73 authorities now published signed "network status" documents. Each
74 network status listed, for every router in the network: a hash of its
75 identity key, a hash of its most recent descriptor, and a summary of
76 what the authority believed about its status. Clients would download
77 the authorities' network status documents in turn, and believe
78 statements about routers iff they were attested to by more than half of
79 the authorities.
81 Instead of downloading all router descriptors at once, clients
82 downloaded only the descriptors that they did not have. Descriptors
83 were indexed by their digests, in order to prevent malicious caches
84 from giving different versions of a router descriptor to different
85 clients.
87 Routers began working harder to upload new descriptors only when their
88 contents were substantially changed.
91 0.2. Goals of the version 3 protocol
93 Version 3 of the Tor directory protocol tries to solve the following
94 issues:
96 * A great deal of bandwidth used to transmit router descriptors was
97 used by two fields that are not actually used by Tor routers
98 (namely read-history and write-history). We save about 60% by
99 moving them into a separate document that most clients do not
100 fetch or use.
102 * It was possible under certain perverse circumstances for clients
103 to download an unusual set of network status documents, thus
104 partitioning themselves from clients who have a more recent and/or
105 typical set of documents. Even under the best of circumstances,
106 clients were sensitive to the ages of the network status documents
107 they downloaded. Therefore, instead of having the clients
108 correlate multiple network status documents, we have the
109 authorities collectively vote on a single consensus network status
110 document.
112 * The most sensitive data in the entire network (the identity keys
113 of the directory authorities) needed to be stored unencrypted so
114 that the authorities can sign network-status documents on the fly.
115 Now, the authorities' identity keys are stored offline, and used
116 to certify medium-term signing keys that can be rotated.
118 0.3. Some Remaining questions
120 Things we could solve on a v3 timeframe:
122 The SHA-1 hash is showing its age. We should do something about our
123 dependency on it. We could probably future-proof ourselves here in
124 this revision, at least so far as documents from the authorities are
125 concerned.
127 Too many things about the authorities are hardcoded by IP.
129 Perhaps we should start accepting longer identity keys for routers
130 too.
132 Things to solve eventually:
134 Requiring every client to know about every router won't scale forever.
136 Requiring every directory cache to know every router won't scale
137 forever.
140 1. Outline
142 There is a small set (say, around 5-10) of semi-trusted directory
143 authorities. A default list of authorities is shipped with the Tor
144 software. Users can change this list, but are encouraged not to do so,
145 in order to avoid partitioning attacks.
147 Every authority has a very-secret, long-term "Authority Identity Key".
148 This is stored encrypted and/or offline, and is used to sign "key
149 certificate" documents. Every key certificate contains a medium-term
150 (3-12 months) "authority signing key", that is used by the authority to
151 sign other directory information. (Note that the authority identity
152 key is distinct from the router identity key that the authority uses
153 in its role as an ordinary router.)
155 Routers periodically upload signed "routers descriptors" to the
156 directory authorities describing their keys, capabilities, and other
157 information. Routers may also upload signed "extra info documents"
158 containing information that is not required for the Tor protocol.
159 Directory authorities serve router descriptors indexed by router
160 identity, or by hash of the descriptor.
162 Routers may act as directory caches to reduce load on the directory
163 authorities. They announce this in their descriptors.
165 Periodically, each directory authority generates a view of
166 the current descriptors and status for known routers. They send a
167 signed summary of this view (a "status vote") to the other
168 authorities. The authorities compute the result of this vote, and sign
169 a "consensus status" document containing the result of the vote.
171 Directory caches download, cache, and re-serve consensus documents.
173 Clients, directory caches, and directory authorities all use consensus
174 documents to find out when their list of routers is out-of-date.
175 (Directory authorities also use vote statuses.) If it is, they download
176 any missing router descriptors. Clients download missing descriptors
177 from caches; caches and authorities download from authorities.
178 Descriptors are downloaded by the hash of the descriptor, not by the
179 server's identity key: this prevents servers from attacking clients by
180 giving them descriptors nobody else uses.
182 All directory information is uploaded and downloaded with HTTP.
184 [Authorities also generate and caches also cache documents produced and
185 used by earlier versions of this protocol; see dir-spec-v1.txt and
186 dir-spec-v2.txt for notes on those versions.]
188 1.1. What's different from version 2?
190 Clients used to download multiple network status documents,
191 corresponding roughly to "status votes" above. They would compute the
192 result of the vote on the client side.
194 Authorities used to sign documents using the same private keys they used
195 for their roles as routers. This forced them to keep these extremely
196 sensitive keys in memory unencrypted.
198 All of the information in extra-info documents used to be kept in the
199 main descriptors.
201 1.2. Document meta-format
203 Router descriptors, directories, and running-routers documents all obey the
204 following lightweight extensible information format.
206 The highest level object is a Document, which consists of one or more
207 Items. Every Item begins with a KeywordLine, followed by zero or more
208 Objects. A KeywordLine begins with a Keyword, optionally followed by
209 whitespace and more non-newline characters, and ends with a newline. A
210 Keyword is a sequence of one or more characters in the set [A-Za-z0-9-].
211 An Object is a block of encoded data in pseudo-Open-PGP-style
212 armor. (cf. RFC 2440)
214 More formally:
216 NL = The ascii LF character (hex value 0x0a).
217 Document ::= (Item | NL)+
218 Item ::= KeywordLine Object*
219 KeywordLine ::= Keyword NL | Keyword WS ArgumentChar+ NL
220 Keyword = KeywordChar+
221 KeywordChar ::= 'A' ... 'Z' | 'a' ... 'z' | '0' ... '9' | '-'
222 ArgumentChar ::= any printing ASCII character except NL.
223 WS = (SP | TAB)+
224 Object ::= BeginLine Base-64-encoded-data EndLine
225 BeginLine ::= "-----BEGIN " Keyword "-----" NL
226 EndLine ::= "-----END " Keyword "-----" NL
228 The BeginLine and EndLine of an Object must use the same keyword.
230 When interpreting a Document, software MUST ignore any KeywordLine that
231 starts with a keyword it doesn't recognize; future implementations MUST NOT
232 require current clients to understand any KeywordLine not currently
233 described.
235 The "opt" keyword was used until Tor 0.1.2.5-alpha for non-critical future
236 extensions. All implementations MUST ignore any item of the form "opt
237 keyword ....." when they would not recognize "keyword ....."; and MUST
238 treat "opt keyword ....." as synonymous with "keyword ......" when keyword
239 is recognized.
241 Implementations before 0.1.2.5-alpha rejected any document with a
242 KeywordLine that started with a keyword that they didn't recognize.
243 When generating documents that need to be read by older versions of Tor,
244 implementations MUST prefix items not recognized by older versions of
245 Tor with an "opt" until those versions of Tor are obsolete. [Note that
246 key certificates, status vote documents, extra info documents, and
247 status consensus documents will never be read by older versions of Tor.]
249 Other implementations that want to extend Tor's directory format MAY
250 introduce their own items. The keywords for extension items SHOULD start
251 with the characters "x-" or "X-", to guarantee that they will not conflict
252 with keywords used by future versions of Tor.
254 In our document descriptions below, we tag Items with a multiplicity in
255 brackets. Possible tags are:
257 "At start, exactly once": These items MUST occur in every instance of
258 the document type, and MUST appear exactly once, and MUST be the
259 first item in their documents.
261 "Exactly once": These items MUST occur exactly one time in every
262 instance of the document type.
264 "At end, exactly once": These items MUST occur in every instance of
265 the document type, and MUST appear exactly once, and MUST be the
266 last item in their documents.
268 "At most once": These items MAY occur zero or one times in any
269 instance of the document type, but MUST NOT occur more than once.
271 "Any number": These items MAY occur zero, one, or more times in any
272 instance of the document type.
274 "Once or more": These items MUST occur at least once in any instance
275 of the document type, and MAY occur more.
277 1.3. Signing documents
279 Every signable document below is signed in a similar manner, using a
280 given "Initial Item", a final "Signature Item", a digest algorithm, and
281 a signing key.
283 The Initial Item must be the first item in the document.
285 The Signature Item has the following format:
287 <signature item keyword> [arguments] NL SIGNATURE NL
289 The "SIGNATURE" Object contains a signature (using the signing key) of
290 the PKCS1-padded digest of the entire document, taken from the
291 beginning of the Initial item, through the newline after the Signature
292 Item's keyword and its arguments.
294 Unless otherwise, the digest algorithm is SHA-1.
296 All documents are invalid unless signed with the correct signing key.
298 The "Digest" of a document, unless stated otherwise, is its digest *as
299 signed by this signature scheme*.
301 1.4. Voting timeline
303 Every consensus document has a "valid-after" (VA) time, a "fresh-until"
304 (FU) time and a "valid-until" (VU) time. VA MUST precede FU, which MUST
305 in turn precede VU. Times are chosen so that every consensus will be
306 "fresh" until the next consensus becomes valid, and "valid" for a while
307 after. At least 3 consensuses should be valid at any given time.
309 The timeline for a given consensus is as follows:
311 VA-DistSeconds-VoteSeconds: The authorities exchange votes.
313 VA-DistSeconds-VoteSeconds/2: The authorities try to download any
314 votes they don't have.
316 VA-DistSeconds: The authorities calculate the consensus and exchange
317 signatures.
319 VA-DistSeconds/2: The authorities try to download any signatures
320 they don't have.
322 VA: All authorities have a multiply signed consensus.
324 VA ... FU: Caches download the consensus. (Note that since caches have
325 no way of telling what VA and FU are until they have downloaded
326 the consensus, they assume that the present consensus's VA is
327 equal to the previous one's FU, and that its FU is one interval after
328 that.)
330 FU: The consensus is no longer the freshest consensus.
332 FU ... (the current consensus's VU): Clients download the consensus.
333 (See note above: clients guess that the next consensus's FU will be
334 two intervals after the current VA.)
336 VU: The consensus is no longer valid.
338 VoteSeconds and DistSeconds MUST each be at least 20 seconds; FU-VA and
339 VU-FU MUST each be at least 5 minutes.
341 2. Router operation and formats
343 ORs SHOULD generate a new router descriptor and a new extra-info
344 document whenever any of the following events have occurred:
346 - A period of time (18 hrs by default) has passed since the last
347 time a descriptor was generated.
349 - A descriptor field other than bandwidth or uptime has changed.
351 - Bandwidth has changed by a factor of 2 from the last time a
352 descriptor was generated, and at least a given interval of time
353 (20 mins by default) has passed since then.
355 - Its uptime has been reset (by restarting).
357 [XXX this list is incomplete; see router_differences_are_cosmetic()
358 in routerlist.c for others]
360 ORs SHOULD NOT publish a new router descriptor or extra-info document
361 if none of the above events have occurred and not much time has passed
362 (12 hours by default).
364 After generating a descriptor, ORs upload them to every directory
365 authority they know, by posting them (in order) to the URL
367 http://<hostname:port>/tor/
369 2.1. Router descriptor format
371 Router descriptors consist of the following items. For backward
372 compatibility, there should be an extra NL at the end of each router
373 descriptor.
375 In lines that take multiple arguments, extra arguments SHOULD be
376 accepted and ignored. Many of the nonterminals below are defined in
377 section 2.3.
379 "router" nickname address ORPort SOCKSPort DirPort NL
381 [At start, exactly once.]
383 Indicates the beginning of a router descriptor. "nickname" must be a
384 valid router nickname as specified in 2.3. "address" must be an IPv4
385 address in dotted-quad format. The last three numbers indicate the
386 TCP ports at which this OR exposes functionality. ORPort is a port at
387 which this OR accepts TLS connections for the main OR protocol;
388 SOCKSPort is deprecated and should always be 0; and DirPort is the
389 port at which this OR accepts directory-related HTTP connections. If
390 any port is not supported, the value 0 is given instead of a port
391 number. (At least one of DirPort and ORPort SHOULD be set;
392 authorities MAY reject any descriptor with both DirPort and ORPort of
393 0.)
395 "bandwidth" bandwidth-avg bandwidth-burst bandwidth-observed NL
397 [Exactly once]
399 Estimated bandwidth for this router, in bytes per second. The
400 "average" bandwidth is the volume per second that the OR is willing to
401 sustain over long periods; the "burst" bandwidth is the volume that
402 the OR is willing to sustain in very short intervals. The "observed"
403 value is an estimate of the capacity this server can handle. The
404 server remembers the max bandwidth sustained output over any ten
405 second period in the past day, and another sustained input. The
406 "observed" value is the lesser of these two numbers.
408 "platform" string NL
410 [At most once]
412 A human-readable string describing the system on which this OR is
413 running. This MAY include the operating system, and SHOULD include
414 the name and version of the software implementing the Tor protocol.
416 "published" YYYY-MM-DD HH:MM:SS NL
418 [Exactly once]
420 The time, in GMT, when this descriptor (and its corresponding
421 extra-info document if any) was generated.
423 "fingerprint" fingerprint NL
425 [At most once]
427 A fingerprint (a HASH_LEN-byte of asn1 encoded public key, encoded in
428 hex, with a single space after every 4 characters) for this router's
429 identity key. A descriptor is considered invalid (and MUST be
430 rejected) if the fingerprint line does not match the public key.
432 [We didn't start parsing this line until Tor 0.1.0.6-rc; it should
433 be marked with "opt" until earlier versions of Tor are obsolete.]
435 "hibernating" bool NL
437 [At most once]
439 If the value is 1, then the Tor server was hibernating when the
440 descriptor was published, and shouldn't be used to build circuits.
442 [We didn't start parsing this line until Tor 0.1.0.6-rc; it should be
443 marked with "opt" until earlier versions of Tor are obsolete.]
445 "uptime" number NL
447 [At most once]
449 The number of seconds that this OR process has been running.
451 "onion-key" NL a public key in PEM format
453 [Exactly once]
455 This key is used to encrypt EXTEND cells for this OR. The key MUST be
456 accepted for at least 1 week after any new key is published in a
457 subsequent descriptor. It MUST be 1024 bits.
459 "signing-key" NL a public key in PEM format
461 [Exactly once]
463 The OR's long-term identity key. It MUST be 1024 bits.
465 "accept" exitpattern NL
466 "reject" exitpattern NL
468 [Any number]
470 These lines describe an "exit policy": the rules that an OR follows
471 when deciding whether to allow a new stream to a given address. The
472 'exitpattern' syntax is described below. There MUST be at least one
473 such entry. The rules are considered in order; if no rule matches,
474 the address will be accepted. For clarity, the last such entry SHOULD
475 be accept *:* or reject *:*.
477 "router-signature" NL Signature NL
479 [At end, exactly once]
481 The "SIGNATURE" object contains a signature of the PKCS1-padded
482 hash of the entire router descriptor, taken from the beginning of the
483 "router" line, through the newline after the "router-signature" line.
484 The router descriptor is invalid unless the signature is performed
485 with the router's identity key.
487 "contact" info NL
489 [At most once]
491 Describes a way to contact the server's administrator, preferably
492 including an email address and a PGP key fingerprint.
494 "family" names NL
496 [At most once]
498 'Names' is a space-separated list of server nicknames or
499 hexdigests. If two ORs list one another in their "family" entries,
500 then OPs should treat them as a single OR for the purpose of path
501 selection.
503 For example, if node A's descriptor contains "family B", and node B's
504 descriptor contains "family A", then node A and node B should never
505 be used on the same circuit.
507 "read-history" YYYY-MM-DD HH:MM:SS (NSEC s) NUM,NUM,NUM,NUM,NUM... NL
508 [At most once]
509 "write-history" YYYY-MM-DD HH:MM:SS (NSEC s) NUM,NUM,NUM,NUM,NUM... NL
510 [At most once]
512 Declare how much bandwidth the OR has used recently. Usage is divided
513 into intervals of NSEC seconds. The YYYY-MM-DD HH:MM:SS field
514 defines the end of the most recent interval. The numbers are the
515 number of bytes used in the most recent intervals, ordered from
516 oldest to newest.
518 [We didn't start parsing these lines until Tor 0.1.0.6-rc; they should
519 be marked with "opt" until earlier versions of Tor are obsolete.]
521 [See also migration notes in section 2.2.1.]
523 "eventdns" bool NL
525 [At most once]
527 Declare whether this version of Tor is using the newer enhanced
528 dns logic. Versions of Tor with this field set to false SHOULD NOT
529 be used for reverse hostname lookups.
531 [All versions of Tor before 0.1.2.2-alpha should be assumed to have
532 this option set to 0 if it is not present. All Tor versions at
533 0.1.2.2-alpha or later should be assumed to have this option set to
534 1 if it is not present. Until 0.1.2.1-alpha-dev, this option was
535 not generated, even when the new DNS code was in use. Versions of Tor
536 before 0.1.2.1-alpha-dev did not parse this option, so it should be
537 marked "opt". The dnsworker logic has been removed, so this option
538 should not be used by new server code. However, it can still be
539 used, and should still be recognized by new code until Tor 0.1.2.x
540 is obsolete.]
542 "caches-extra-info" NL
544 [At most once.]
546 Present only if this router is a directory cache that provides
547 extra-info documents.
549 [Versions before 0.2.0.1-alpha don't recognize this, and versions
550 before 0.1.2.5-alpha will reject descriptors containing it unless
551 it is prefixed with "opt"; it should be so prefixed until these
552 versions are obsolete.]
554 "extra-info-digest" digest NL
556 [At most once]
558 "Digest" is a hex-encoded digest (using upper-case characters) of the
559 router's extra-info document, as signed in the router's extra-info
560 (that is, not including the signature). (If this field is absent, the
561 router is not uploading a corresponding extra-info document.)
563 [Versions before 0.2.0.1-alpha don't recognize this, and versions
564 before 0.1.2.5-alpha will reject descriptors containing it unless
565 it is prefixed with "opt"; it should be so prefixed until these
566 versions are obsolete.]
568 "hidden-service-dir" *(SP VersionNum) NL
570 [At most once.]
572 Present only if this router stores and serves hidden service
573 descriptors. If any VersionNum(s) are specified, this router
574 supports those descriptor versions. If none are specified, it
575 defaults to version 2 descriptors.
577 [Versions of Tor before 0.1.2.5-alpha rejected router descriptors
578 with unrecognized items; the protocols line should be preceded with
579 an "opt" until these Tors are obsolete.]
581 "protocols" SP "Link" SP LINK-VERSION-LIST SP "Circuit" SP
582 CIRCUIT-VERSION-LIST NL
584 [At most once.]
586 Both lists are space-separated sequences of numbers, to indicate which
587 protocols the server supports. As of 30 Mar 2008, specified
588 protocols are "Link 1 2 Circuit 1". See section 4.1 of tor-spec.txt
589 for more information about link protocol versions.
591 [Versions of Tor before 0.1.2.5-alpha rejected router descriptors
592 with unrecognized items; the protocols line should be preceded with
593 an "opt" until these Tors are obsolete.]
595 "allow-single-hop-exits"
597 [At most once.]
599 Present only if the router allows single-hop circuits to make exit
600 connections. Most Tor servers do not support this: this is
601 included for specialized controllers designed to support perspective
602 access and such.
605 2.2. Extra-info documents
607 Extra-info documents consist of the following items:
609 "extra-info" Nickname Fingerprint NL
610 [At start, exactly once.]
612 Identifies what router this is an extra info descriptor for.
613 Fingerprint is encoded in hex (using upper-case letters), with
614 no spaces.
616 "published"
618 [Exactly once.]
620 The time, in GMT, when this document (and its corresponding router
621 descriptor if any) was generated. It MUST match the published time
622 in the corresponding router descriptor.
624 "read-history" YYYY-MM-DD HH:MM:SS (NSEC s) NUM,NUM,NUM,NUM,NUM... NL
625 [At most once.]
626 "write-history" YYYY-MM-DD HH:MM:SS (NSEC s) NUM,NUM,NUM,NUM,NUM... NL
627 [At most once.]
629 As documented in 2.1 above. See migration notes in section 2.2.1.
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
643 statistics.
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
655 [At most once.]
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
662 least 24 hours.
664 "bridge-ips" CC=N,CC=N,... NL
665 [At most once.]
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
670 multiple of 8.
672 "dirreq-stats-end" YYYY-MM-DD HH:MM:SS (NSEC s) NL
673 [At most once.]
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
683 [At most once.]
684 "dirreq-v3-ips" CC=N,CC=N,... NL
685 [At most once.]
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
694 [At most once.]
695 "dirreq-v3-reqs" CC=N,CC=N,... NL
696 [At most once.]
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
704 [At most once.]
705 "dirreq-v3-share" num% NL
706 [At most once.]
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
715 [At most once.]
716 "dirreq-v3-resp" status=nul,... NL
717 [At most once.]
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
729 rounding up.
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
739 [At most once.]
740 "dirreq-v3-direct-dl" key=val,... NL
741 [At most once.]
742 "dirreq-v2-tunneled-dl" key=val,... NL
743 [At most once.]
744 "dirreq-v3-tunneled-dl" key=val,... NL
745 [At most once.]
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
753 as follows:
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
762 send the response.
764 Download times:
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
775 q3.
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
778 bandwidth than md.
780 "entry-stats-end" YYYY-MM-DD HH:MM:SS (NSEC s) NL
781 [At most once.]
783 YYYY-MM-DD HH:MM:SS defines the end of the included measurement
784 interval of length NSEC seconds (86400 seconds by default).
786 An "entry-stats-end" line, as well as any other "entry-*"
787 line, is first added after the relay has been running for at least
788 24 hours.
790 "entry-ips" CC=N,CC=N,... NL
791 [At most once.]
793 List of mappings from two-letter country codes to the number of
794 unique IP addresses that have connected from that country to the
795 relay and which are no known other relays, rounded up to the
796 nearest multiple of 8.
798 "cell-stats-end" YYYY-MM-DD HH:MM:SS (NSEC s) NL
799 [At most once.]
801 YYYY-MM-DD HH:MM:SS defines the end of the included measurement
802 interval of length NSEC seconds (86400 seconds by default).
804 A "cell-stats-end" line, as well as any other "cell-*" line,
805 is first added after the relay has been running for at least 24
806 hours.
808 "cell-processed-cells" num,...,num NL
809 [At most once.]
811 Mean number of processed cells per circuit, subdivided into
812 deciles of circuits by the number of cells they have processed in
813 descending order from loudest to quietest circuits.
815 "cell-queued-cells" num,...,num NL
816 [At most once.]
818 Mean number of cells contained in queues by circuit decile. These
819 means are calculated by 1) determining the mean number of cells in
820 a single circuit between its creation and its termination and 2)
821 calculating the mean for all circuits in a given decile as
822 determined in "cell-processed-cells". Numbers have a precision of
823 two decimal places.
825 "cell-time-in-queue" num,...,num NL
826 [At most once.]
828 Mean time cells spend in circuit queues in milliseconds. Times are
829 calculated by 1) determining the mean time cells spend in the
830 queue of a single circuit and 2) calculating the mean for all
831 circuits in a given decile as determined in
832 "cell-processed-cells".
834 "cell-circuits-per-decile" num NL
835 [At most once.]
837 Mean number of circuits that are included in any of the deciles,
838 rounded up to the next integer.
840 "exit-stats-end" YYYY-MM-DD HH:MM:SS (NSEC s) NL
841 [At most once.]
843 YYYY-MM-DD HH:MM:SS defines the end of the included measurement
844 interval of length NSEC seconds (86400 seconds by default).
846 An "exit-stats-end" line, as well as any other "exit-*" line, is
847 first added after the relay has been running for at least 24 hours
848 and only if the relay permits exiting (where exiting to a single
849 port and IP address is sufficient).
851 "exit-kibibytes-written" port=N,port=N,... NL
852 [At most once.]
853 "exit-kibibytes-read" port=N,port=N,... NL
854 [At most once.]
856 List of mappings from ports to the number of kibibytes that the
857 relay has written to or read from exit connections to that port,
858 rounded up to the next full kibibyte.
860 "exit-streams-opened" port=N,port=N,... NL
861 [At most once.]
863 List of mappings from ports to the number of opened exit streams
864 to that port, rounded up to the nearest multiple of 4.
866 "router-signature" NL Signature NL
867 [At end, exactly once.]
869 A document signature as documented in section 1.3, using the
870 initial item "extra-info" and the final item "router-signature",
871 signed with the router's identity key.
873 2.2.1. Moving history fields to extra-info documents.
875 Tools that want to use the read-history and write-history values SHOULD
876 download extra-info documents as well as router descriptors. Such
877 tools SHOULD accept history values from both sources; if they appear in
878 both documents, the values in the extra-info documents are authoritative.
880 New versions of Tor no longer generate router descriptors
881 containing read-history or write-history. Tools should continue to
882 accept read-history and write-history values in router descriptors
883 produced by older versions of Tor until all Tor versions earlier
884 than 0.2.0.x are obsolete.
886 2.3. Nonterminals in router descriptors
888 nickname ::= between 1 and 19 alphanumeric characters ([A-Za-z0-9]),
889 case-insensitive.
890 hexdigest ::= a '$', followed by 40 hexadecimal characters
891 ([A-Fa-f0-9]). [Represents a server by the digest of its identity
892 key.]
894 exitpattern ::= addrspec ":" portspec
895 portspec ::= "*" | port | port "-" port
896 port ::= an integer between 1 and 65535, inclusive.
898 [Some implementations incorrectly generate ports with value 0.
899 Implementations SHOULD accept this, and SHOULD NOT generate it.
900 Connections to port 0 are never permitted.]
902 addrspec ::= "*" | ip4spec | ip6spec
903 ipv4spec ::= ip4 | ip4 "/" num_ip4_bits | ip4 "/" ip4mask
904 ip4 ::= an IPv4 address in dotted-quad format
905 ip4mask ::= an IPv4 mask in dotted-quad format
906 num_ip4_bits ::= an integer between 0 and 32
907 ip6spec ::= ip6 | ip6 "/" num_ip6_bits
908 ip6 ::= an IPv6 address, surrounded by square brackets.
909 num_ip6_bits ::= an integer between 0 and 128
911 bool ::= "0" | "1"
913 3. Formats produced by directory authorities.
915 Every authority has two keys used in this protocol: a signing key, and
916 an authority identity key. (Authorities also have a router identity
917 key used in their role as a router and by earlier versions of the
918 directory protocol.) The identity key is used from time to time to
919 sign new key certificates using new signing keys; it is very sensitive.
920 The signing key is used to sign key certificates and status documents.
922 There are three kinds of documents generated by directory authorities:
924 Key certificates
925 Status votes
926 Status consensuses
928 Each is discussed below.
930 3.1. Key certificates
932 Key certificates consist of the following items:
934 "dir-key-certificate-version" version NL
936 [At start, exactly once.]
938 Determines the version of the key certificate. MUST be "3" for
939 the protocol described in this document. Implementations MUST
940 reject formats they don't understand.
942 "dir-address" IPPort NL
943 [At most once]
945 An IP:Port for this authority's directory port.
947 "fingerprint" fingerprint NL
949 [Exactly once.]
951 Hexadecimal encoding without spaces based on the authority's
952 identity key.
954 "dir-identity-key" NL a public key in PEM format
956 [Exactly once.]
958 The long-term authority identity key for this authority. This key
959 SHOULD be at least 2048 bits long; it MUST NOT be shorter than
960 1024 bits.
962 "dir-key-published" YYYY-MM-DD HH:MM:SS NL
964 [Exactly once.]
966 The time (in GMT) when this document and corresponding key were
967 last generated.
969 "dir-key-expires" YYYY-MM-DD HH:MM:SS NL
971 [Exactly once.]
973 A time (in GMT) after which this key is no longer valid.
975 "dir-signing-key" NL a key in PEM format
977 [Exactly once.]
979 The directory server's public signing key. This key MUST be at
980 least 1024 bits, and MAY be longer.
982 "dir-key-crosscert" NL CrossSignature NL
984 [At most once.]
986 NOTE: Authorities MUST include this field in all newly generated
987 certificates. A future version of this specification will make
988 the field required.
990 CrossSignature is a signature, made using the certificate's signing
991 key, of the digest of the PKCS1-padded hash of the certificate's
992 identity key. For backward compatibility with broken versions of the
993 parser, we wrap the base64-encoded signature in -----BEGIN ID
994 SIGNATURE---- and -----END ID SIGNATURE----- tags. Implementations
995 MUST allow the "ID " portion to be omitted, however.
997 When encountering a certificate with a dir-key-crosscert entry,
998 implementations MUST verify that the signature is a correct signature
999 of the hash of the identity key using the signing key.
1001 "dir-key-certification" NL Signature NL
1003 [At end, exactly once.]
1005 A document signature as documented in section 1.3, using the
1006 initial item "dir-key-certificate-version" and the final item
1007 "dir-key-certification", signed with the authority identity key.
1009 Authorities MUST generate a new signing key and corresponding
1010 certificate before the key expires.
1012 3.2. Vote and consensus status documents
1014 Votes and consensuses are more strictly formatted then other documents
1015 in this specification, since different authorities must be able to
1016 generate exactly the same consensus given the same set of votes.
1018 The procedure for deciding when to generate vote and consensus status
1019 documents are described in section 1.4 on the voting timeline.
1021 Status documents contain a preamble, an authority section, a list of
1022 router status entries, and one or more footer signature, in that order.
1024 Unlike other formats described above, a SP in these documents must be a
1025 single space character (hex 20).
1027 Some items appear only in votes, and some items appear only in
1028 consensuses. Unless specified, items occur in both.
1030 The preamble contains the following items. They MUST occur in the
1031 order given here:
1033 "network-status-version" SP version NL.
1035 [At start, exactly once.]
1037 A document format version. For this specification, the version is
1038 "3".
1040 "vote-status" SP type NL
1042 [Exactly once.]
1044 The status MUST be "vote" or "consensus", depending on the type of
1045 the document.
1047 "consensus-methods" SP IntegerList NL
1049 [Exactly once for votes; does not occur in consensuses.]
1051 A space-separated list of supported methods for generating
1052 consensuses from votes. See section 3.4.1 for details. Method "1"
1053 MUST be included.
1055 "consensus-method" SP Integer NL
1057 [Exactly once for consensuses; does not occur in votes.]
1059 See section 3.4.1 for details.
1061 (Only included when the vote is generated with consensus-method 2 or
1062 later.)
1064 "published" SP YYYY-MM-DD SP HH:MM:SS NL
1066 [Exactly once for votes; does not occur in consensuses.]
1068 The publication time for this status document (if a vote).
1070 "valid-after" SP YYYY-MM-DD SP HH:MM:SS NL
1072 [Exactly once.]
1074 The start of the Interval for this vote. Before this time, the
1075 consensus document produced from this vote should not be used.
1076 See 1.4 for voting timeline information.
1078 "fresh-until" SP YYYY-MM-DD SP HH:MM:SS NL
1080 [Exactly once.]
1082 The time at which the next consensus should be produced; before this
1083 time, there is no point in downloading another consensus, since there
1084 won't be a new one. See 1.4 for voting timeline information.
1086 "valid-until" SP YYYY-MM-DD SP HH:MM:SS NL
1088 [Exactly once.]
1090 The end of the Interval for this vote. After this time, the
1091 consensus produced by this vote should not be used. See 1.4 for
1092 voting timeline information.
1094 "voting-delay" SP VoteSeconds SP DistSeconds NL
1096 [Exactly once.]
1098 VoteSeconds is the number of seconds that we will allow to collect
1099 votes from all authorities; DistSeconds is the number of seconds
1100 we'll allow to collect signatures from all authorities. See 1.4 for
1101 voting timeline information.
1103 "client-versions" SP VersionList NL
1105 [At most once.]
1107 A comma-separated list of recommended client versions, in
1108 ascending order. If absent, no opinion is held about client
1109 versions.
1111 "server-versions" SP VersionList NL
1113 [At most once.]
1115 A comma-separated list of recommended server versions, in
1116 ascending order. If absent, no opinion is held about server
1117 versions.
1119 "known-flags" SP FlagList NL
1121 [Exactly once.]
1123 A space-separated list of all of the flags that this document
1124 might contain. A flag is "known" either because the authority
1125 knows about them and might set them (if in a vote), or because
1126 enough votes were counted for the consensus for an authoritative
1127 opinion to have been formed about their status.
1129 "params" SP [Parameters] NL
1131 [At most once]
1133 Parameter ::= Keyword '=' Int32
1134 Int32 ::= A decimal integer between -2147483648 and 2147483647.
1135 Parameters ::= Parameter | Parameters SP Parameter
1137 The parameters list, if present, contains a space-separated list of
1138 key-value pairs, sorted in lexical order by their keyword. Each
1139 parameter has its own meaning.
1141 (Only included when the vote is generated with consensus-method 7 or
1142 later.)
1144 Commonly used "param" arguments at this point include:
1146 "CircWindow" -- the default package window that circuits should
1147 be established with. It started out at 1000 cells, but some
1148 research indicates that a lower value would mean fewer cells in
1149 transit in the network at any given time. Obeyed by Tor 0.2.1.20
1150 and later.
1152 "CircPriorityHalflifeMsec" -- the halflife parameter used when
1153 weighting which circuit will send the next cell. Obeyed by Tor
1154 0.2.2.7-alpha and later.
1156 The authority section of a vote contains the following items, followed
1157 in turn by the authority's current key certificate:
1159 "dir-source" SP nickname SP identity SP address SP IP SP dirport SP
1160 orport NL
1162 [Exactly once, at start]
1164 Describes this authority. The nickname is a convenient identifier
1165 for the authority. The identity is an uppercase hex fingerprint of
1166 the authority's current (v3 authority) identity key. The address is
1167 the server's hostname. The IP is the server's current IP address,
1168 and dirport is its current directory port. XXXXorport
1170 "contact" SP string NL
1172 [At most once.]
1174 An arbitrary string describing how to contact the directory
1175 server's administrator. Administrators should include at least an
1176 email address and a PGP fingerprint.
1178 "legacy-key" SP FINGERPRINT NL
1180 [At most once]
1182 Lists a fingerprint for an obsolete _identity_ key still used
1183 by this authority to keep older clients working. This option
1184 is used to keep key around for a little while in case the
1185 authorities need to migrate many identity keys at once.
1186 (Generally, this would only happen because of a security
1187 vulnerability that affected multiple authorities, like the
1188 Debian OpenSSL RNG bug of May 2008.)
1190 The authority section of a consensus contains groups the following items,
1191 in the order given, with one group for each authority that contributed to
1192 the consensus, with groups sorted by authority identity digest:
1194 "dir-source" SP nickname SP identity SP address SP IP SP dirport SP
1195 orport NL
1197 [Exactly once, at start]
1199 As in the authority section of a vote.
1201 "contact" SP string NL
1203 [At most once.]
1205 As in the authority section of a vote.
1207 "vote-digest" SP digest NL
1209 [Exactly once.]
1211 A digest of the vote from the authority that contributed to this
1212 consensus, as signed (that is, not including the signature).
1213 (Hex, upper-case.)
1215 Each router status entry contains the following items. Router status
1216 entries are sorted in ascending order by identity digest.
1218 "r" SP nickname SP identity SP digest SP publication SP IP SP ORPort
1219 SP DirPort NL
1221 [At start, exactly once.]
1223 "Nickname" is the OR's nickname. "Identity" is a hash of its
1224 identity key, encoded in base64, with trailing equals sign(s)
1225 removed. "Digest" is a hash of its most recent descriptor as
1226 signed (that is, not including the signature), encoded in base64.
1227 "Publication" is the
1228 publication time of its most recent descriptor, in the form
1229 YYYY-MM-DD HH:MM:SS, in GMT. "IP" is its current IP address;
1230 ORPort is its current OR port, "DirPort" is it's current directory
1231 port, or "0" for "none".
1233 "s" SP Flags NL
1235 [At most once.]
1237 A series of space-separated status flags, in alphabetical order.
1238 Currently documented flags are:
1240 "Authority" if the router is a directory authority.
1241 "BadExit" if the router is believed to be useless as an exit node
1242 (because its ISP censors it, because it is behind a restrictive
1243 proxy, or for some similar reason).
1244 "BadDirectory" if the router is believed to be useless as a
1245 directory cache (because its directory port isn't working,
1246 its bandwidth is always throttled, or for some similar
1247 reason).
1248 "Exit" if the router is more useful for building
1249 general-purpose exit circuits than for relay circuits. The
1250 path building algorithm uses this flag; see path-spec.txt.
1251 "Fast" if the router is suitable for high-bandwidth circuits.
1252 "Guard" if the router is suitable for use as an entry guard.
1253 "HSDir" if the router is considered a v2 hidden service directory.
1254 "Named" if the router's identity-nickname mapping is canonical,
1255 and this authority binds names.
1256 "Stable" if the router is suitable for long-lived circuits.
1257 "Running" if the router is currently usable.
1258 "Unnamed" if another router has bound the name used by this
1259 router, and this authority binds names.
1260 "Valid" if the router has been 'validated'.
1261 "V2Dir" if the router implements the v2 directory protocol.
1262 "V3Dir" if the router implements this protocol.
1264 "v" SP version NL
1266 [At most once.]
1268 The version of the Tor protocol that this server is running. If
1269 the value begins with "Tor" SP, the rest of the string is a Tor
1270 version number, and the protocol is "The Tor protocol as supported
1271 by the given version of Tor." Otherwise, if the value begins with
1272 some other string, Tor has upgraded to a more sophisticated
1273 protocol versioning system, and the protocol is "a version of the
1274 Tor protocol more recent than any we recognize."
1276 Directory authorities SHOULD omit version strings they receive from
1277 descriptors if they would cause "v" lines to be over 128 characters
1278 long.
1280 "w" SP "Bandwidth=" INT [SP "Measured=" INT] NL
1282 [At most once.]
1284 An estimate of the bandwidth of this server, in an arbitrary
1285 unit (currently kilobytes per second). Used to weight router
1286 selection.
1288 Additionally, the Measured= keyword is present in votes by
1289 participating bandwidth measurement authorities to indicate
1290 a measured bandwidth currently produced by measuring stream
1291 capacities.
1293 Other weighting keywords may be added later.
1294 Clients MUST ignore keywords they do not recognize.
1296 "p" SP ("accept" / "reject") SP PortList NL
1298 [At most once.]
1300 PortList = PortOrRange
1301 PortList = PortList "," PortOrRange
1302 PortOrRange = INT "-" INT / INT
1304 A list of those ports that this router supports (if 'accept')
1305 or does not support (if 'reject') for exit to "most
1306 addresses".
1308 The footer section is delineated in all votes and consensuses supporting
1309 consensus method 9 and above with the following:
1311 "directory-footer" NL
1313 It contains two subsections, a bandwidths-weights line and a
1314 directory-signature.
1316 The bandwidths-weights line appears At Most Once for a consensus. It does
1317 not appear in votes.
1319 "bandwidth-weights" SP
1320 "Wbd=" INT SP "Wbe=" INT SP "Wbg=" INT SP "Wbm=" INT SP
1321 "Wdb=" INT SP
1322 "Web=" INT SP "Wed=" INT SP "Wee=" INT SP "Weg=" INT SP "Wem=" INT SP
1323 "Wgb=" INT SP "Wgd=" INT SP "Wgg=" INT SP "Wgm=" INT SP
1324 "Wmb=" INT SP "Wmd=" INT SP "Wme=" INT SP "Wmg=" INT SP "Wmm=" INT NL
1326 These values represent the weights to apply to router bandwidths during
1327 path selection. They are sorted in alphabetical order in the list. The
1328 integer values are divided by BW_WEIGHT_SCALE=10000 or the consensus
1329 param "bwweightscale". They are:
1331 Wgg - Weight for Guard-flagged nodes in the guard position
1332 Wgm - Weight for non-flagged nodes in the guard Position
1333 Wgd - Weight for Guard+Exit-flagged nodes in the guard Position
1335 Wmg - Weight for Guard-flagged nodes in the middle Position
1336 Wmm - Weight for non-flagged nodes in the middle Position
1337 Wme - Weight for Exit-flagged nodes in the middle Position
1338 Wmd - Weight for Guard+Exit flagged nodes in the middle Position
1340 Weg - Weight for Guard flagged nodes in the exit Position
1341 Wem - Weight for non-flagged nodes in the exit Position
1342 Wee - Weight for Exit-flagged nodes in the exit Position
1343 Wed - Weight for Guard+Exit-flagged nodes in the exit Position
1345 Wgb - Weight for BEGIN_DIR-supporting Guard-flagged nodes
1346 Wmb - Weight for BEGIN_DIR-supporting non-flagged nodes
1347 Web - Weight for BEGIN_DIR-supporting Exit-flagged nodes
1348 Wdb - Weight for BEGIN_DIR-supporting Guard+Exit-flagged nodes
1350 Wbg - Weight for Guard+Exit-flagged nodes for BEGIN_DIR requests
1351 Wbm - Weight for Guard+Exit-flagged nodes for BEGIN_DIR requests
1352 Wbe - Weight for Guard+Exit-flagged nodes for BEGIN_DIR requests
1353 Wbd - Weight for Guard+Exit-flagged nodes for BEGIN_DIR requests
1355 These values are calculated as specified in Section 3.4.3.
1357 The signature contains the following item, which appears Exactly Once
1358 for a vote, and At Least Once for a consensus.
1360 "directory-signature" SP identity SP signing-key-digest NL Signature
1362 This is a signature of the status document, with the initial item
1363 "network-status-version", and the signature item
1364 "directory-signature", using the signing key. (In this case, we take
1365 the hash through the _space_ after directory-signature, not the
1366 newline: this ensures that all authorities sign the same thing.)
1367 "identity" is the hex-encoded digest of the authority identity key of
1368 the signing authority, and "signing-key-digest" is the hex-encoded
1369 digest of the current authority signing key of the signing authority.
1371 3.3. Assigning flags in a vote
1373 (This section describes how directory authorities choose which status
1374 flags to apply to routers, as of Tor 0.2.0.0-alpha-dev. Later directory
1375 authorities MAY do things differently, so long as clients keep working
1376 well. Clients MUST NOT depend on the exact behaviors in this section.)
1378 In the below definitions, a router is considered "active" if it is
1379 running, valid, and not hibernating.
1381 "Valid" -- a router is 'Valid' if it is running a version of Tor not
1382 known to be broken, and the directory authority has not blacklisted
1383 it as suspicious.
1385 "Named" -- Directory authority administrators may decide to support name
1386 binding. If they do, then they must maintain a file of
1387 nickname-to-identity-key mappings, and try to keep this file consistent
1388 with other directory authorities. If they don't, they act as clients, and
1389 report bindings made by other directory authorities (name X is bound to
1390 identity Y if at least one binding directory lists it, and no directory
1391 binds X to some other Y'.) A router is called 'Named' if the router
1392 believes the given name should be bound to the given key.
1394 Two strategies exist on the current network for deciding on
1395 values for the Named flag. In the original version, server
1396 operators were asked to send nickname-identity pairs to a
1397 mailing list of Naming directory authorities operators. The
1398 operators were then supposed to add the pairs to their
1399 mapping files; in practice, they didn't get to this often.
1401 Newer Naming authorities run a script that registers routers
1402 in their mapping files once the routers have been online at
1403 least two weeks, no other router has that nickname, and no
1404 other router has wanted the nickname for a month. If a router
1405 has not been online for six months, the router is removed.
1407 "Unnamed" -- Directory authorities that support naming should vote for a
1408 router to be 'Unnamed' if its given nickname is mapped to a different
1409 identity.
1411 "Running" -- A router is 'Running' if the authority managed to connect to
1412 it successfully within the last 30 minutes.
1414 "Stable" -- A router is 'Stable' if it is active, and either its Weighted
1415 MTBF is at least the median for known active routers or its Weighted MTBF
1416 corresponds to at least 7 days. Routers are never called Stable if they are
1417 running a version of Tor known to drop circuits stupidly. (0.1.1.10-alpha
1418 through 0.1.1.16-rc are stupid this way.)
1420 To calculate weighted MTBF, compute the weighted mean of the lengths
1421 of all intervals when the router was observed to be up, weighting
1422 intervals by $\alpha^n$, where $n$ is the amount of time that has
1423 passed since the interval ended, and $\alpha$ is chosen so that
1424 measurements over approximately one month old no longer influence the
1425 weighted MTBF much.
1427 [XXXX what happens when we have less than 4 days of MTBF info.]
1429 "Exit" -- A router is called an 'Exit' iff it allows exits to at
1430 least two of the ports 80, 443, and 6667 and allows exits to at
1431 least one /8 address space.
1433 "Fast" -- A router is 'Fast' if it is active, and its bandwidth is
1434 either in the top 7/8ths for known active routers or at least 20KB/s.
1436 "Guard" -- A router is a possible 'Guard' if its Weighted Fractional
1437 Uptime is at least the median for "familiar" active routers, and if
1438 its bandwidth is at least median or at least 250KB/s.
1439 If the total bandwidth of active non-BadExit Exit servers is less
1440 than one third of the total bandwidth of all active servers, no Exit is
1441 listed as a Guard.
1443 To calculate weighted fractional uptime, compute the fraction
1444 of time that the router is up in any given day, weighting so that
1445 downtime and uptime in the past counts less.
1447 A node is 'familiar' if 1/8 of all active nodes have appeared more
1448 recently than it, OR it has been around for a few weeks.
1450 "Authority" -- A router is called an 'Authority' if the authority
1451 generating the network-status document believes it is an authority.
1453 "V2Dir" -- A router supports the v2 directory protocol if it has an open
1454 directory port, and it is running a version of the directory protocol that
1455 supports the functionality clients need. (Currently, this is
1456 0.1.1.9-alpha or later.)
1458 "V3Dir" -- A router supports the v3 directory protocol if it has an open
1459 directory port, and it is running a version of the directory protocol that
1460 supports the functionality clients need. (Currently, this is
1461 0.2.0.?????-alpha or later.)
1463 "HSDir" -- A router is a v2 hidden service directory if it stores and
1464 serves v2 hidden service descriptors and the authority managed to connect
1465 to it successfully within the last 24 hours.
1467 Directory server administrators may label some servers or IPs as
1468 blacklisted, and elect not to include them in their network-status lists.
1470 Authorities SHOULD 'disable' any servers in excess of 3 on any single IP.
1471 When there are more than 3 to choose from, authorities should first prefer
1472 authorities to non-authorities, then prefer Running to non-Running, and
1473 then prefer high-bandwidth to low-bandwidth. To 'disable' a server, the
1474 authority *should* advertise it without the Running or Valid flag.
1476 Thus, the network-status vote includes all non-blacklisted,
1477 non-expired, non-superseded descriptors.
1479 The bandwidth in a "w" line should be taken as the best estimate
1480 of the router's actual capacity that the authority has. For now,
1481 this should be the lesser of the observed bandwidth and bandwidth
1482 rate limit from the router descriptor. It is given in kilobytes
1483 per second, and capped at some arbitrary value (currently 10 MB/s).
1485 The Measured= keyword on a "w" line vote is currently computed
1486 by multiplying the previous published consensus bandwidth by the
1487 ratio of the measured average node stream capacity to the network
1488 average. If 3 or more authorities provide a Measured= keyword for
1489 a router, the authorities produce a consensus containing a "w"
1490 Bandwidth= keyword equal to the median of the Measured= votes.
1492 The ports listed in a "p" line should be taken as those ports for
1493 which the router's exit policy permits 'most' addresses, ignoring any
1494 accept not for all addresses, ignoring all rejects for private
1495 netblocks. "Most" addresses are permitted if no more than 2^25
1496 IPv4 addresses (two /8 networks) were blocked. The list is encoded
1497 as described in 3.4.2.
1499 3.4. Computing a consensus from a set of votes
1501 Given a set of votes, authorities compute the contents of the consensus
1502 document as follows:
1504 The "valid-after", "valid-until", and "fresh-until" times are taken as
1505 the median of the respective values from all the votes.
1507 The times in the "voting-delay" line are taken as the median of the
1508 VoteSeconds and DistSeconds times in the votes.
1510 Known-flags is the union of all flags known by any voter.
1512 Entries are given on the "params" line for every keyword on which any
1513 authority voted. The values given are the low-median of all votes on
1514 that keyword.
1516 "client-versions" and "server-versions" are sorted in ascending
1517 order; A version is recommended in the consensus if it is recommended
1518 by more than half of the voting authorities that included a
1519 client-versions or server-versions lines in their votes.
1521 The authority item groups (dir-source, contact, fingerprint,
1522 vote-digest) are taken from the votes of the voting
1523 authorities. These groups are sorted by the digests of the
1524 authorities identity keys, in ascending order. If the consensus
1525 method is 3 or later, a dir-source line must be included for
1526 every vote with legacy-key entry, using the legacy-key's
1527 fingerprint, the voter's ordinary nickname with the string
1528 "-legacy" appended, and all other fields as from the original
1529 vote's dir-source line.
1531 A router status entry:
1532 * is included in the result if some router status entry with the same
1533 identity is included by more than half of the authorities (total
1534 authorities, not just those whose votes we have).
1536 * For any given identity, we include at most one router status entry.
1538 * A router entry has a flag set if that is included by more than half
1539 of the authorities who care about that flag.
1541 * Two router entries are "the same" if they have the same
1542 <descriptor digest, published time, nickname, IP, ports> tuple.
1543 We choose the tuple for a given router as whichever tuple appears
1544 for that router in the most votes. We break ties first in favor of
1545 the more recently published, then in favor of smaller server
1546 descriptor digest.
1548 * The Named flag appears if it is included for this routerstatus by
1549 _any_ authority, and if all authorities that list it list the same
1550 nickname. However, if consensus-method 2 or later is in use, and
1551 any authority calls this identity/nickname pair Unnamed, then
1552 this routerstatus does not get the Named flag.
1554 * If consensus-method 2 or later is in use, the Unnamed flag is
1555 set for a routerstatus if any authorities have voted for a different
1556 identities to be Named with that nickname, or if any authority
1557 lists that nickname/ID pair as Unnamed.
1559 (With consensus-method 1, Unnamed is set like any other flag.)
1561 * The version is given as whichever version is listed by the most
1562 voters, with ties decided in favor of more recent versions.
1564 * If consensus-method 4 or later is in use, then routers that
1565 do not have the Running flag are not listed at all.
1567 * If consensus-method 5 or later is in use, then the "w" line
1568 is generated using a low-median of the bandwidth values from
1569 the votes that included "w" lines for this router.
1571 * If consensus-method 5 or later is in use, then the "p" line
1572 is taken from the votes that have the same policy summary
1573 for the descriptor we are listing. (They should all be the
1574 same. If they are not, we pick the most commonly listed
1575 one, breaking ties in favor of the lexicographically larger
1576 vote.) The port list is encoded as specified in 3.4.2.
1578 * If consensus-method 6 or later is in use and if 3 or more
1579 authorities provide a Measured= keyword in their votes for
1580 a router, the authorities produce a consensus containing a
1581 Bandwidth= keyword equal to the median of the Measured= votes.
1583 * If consensus-method 7 or later is in use, the params line is
1584 included in the output.
1586 The signatures at the end of a consensus document are sorted in
1587 ascending order by identity digest.
1589 All ties in computing medians are broken in favor of the smaller or
1590 earlier item.
1592 3.4.1. Forward compatibility
1594 Future versions of Tor will need to include new information in the
1595 consensus documents, but it is important that all authorities (or at least
1596 half) generate and sign the same signed consensus.
1598 To achieve this, authorities list in their votes their supported methods
1599 for generating consensuses from votes. Later methods will be assigned
1600 higher numbers. Currently recognized methods:
1601 "1" -- The first implemented version.
1602 "2" -- Added support for the Unnamed flag.
1603 "3" -- Added legacy ID key support to aid in authority ID key rollovers
1604 "4" -- No longer list routers that are not running in the consensus
1605 "5" -- adds support for "w" and "p" lines.
1606 "6" -- Prefers measured bandwidth values rather than advertised
1607 "7" -- Provides keyword=integer pairs of consensus parameters
1608 "8" -- Provides microdescriptor summaries
1609 "9" -- Provides weights for selecting flagged routers in paths
1611 Before generating a consensus, an authority must decide which consensus
1612 method to use. To do this, it looks for the highest version number
1613 supported by more than 2/3 of the authorities voting. If it supports this
1614 method, then it uses it. Otherwise, it falls back to method 1.
1616 (The consensuses generated by new methods must be parsable by
1617 implementations that only understand the old methods, and must not cause
1618 those implementations to compromise their anonymity. This is a means for
1619 making changes in the contents of consensus; not for making
1620 backward-incompatible changes in their format.)
1622 3.4.2. Encoding port lists
1624 Whether the summary shows the list of accepted ports or the list of
1625 rejected ports depends on which list is shorter (has a shorter string
1626 representation). In case of ties we choose the list of accepted
1627 ports. As an exception to this rule an allow-all policy is
1628 represented as "accept 1-65535" instead of "reject " and a reject-all
1629 policy is similarly given as "reject 1-65535".
1631 Summary items are compressed, that is instead of "80-88,89-100" there
1632 only is a single item of "80-100", similarly instead of "20,21" a
1633 summary will say "20-21".
1635 Port lists are sorted in ascending order.
1637 The maximum allowed length of a policy summary (including the "accept "
1638 or "reject ") is 1000 characters. If a summary exceeds that length we
1639 use an accept-style summary and list as much of the port list as is
1640 possible within these 1000 bytes. [XXXX be more specific.]
1642 3.4.3. Computing Bandwidth Weights
1644 Let weight_scale = 10000
1646 Let G be the total bandwidth for Guard-flagged nodes.
1647 Let M be the total bandwidth for non-flagged nodes.
1648 Let E be the total bandwidth for Exit-flagged nodes.
1649 Let D be the total bandwidth for Guard+Exit-flagged nodes.
1650 Let T = G+M+E+D
1652 Let Wgd be the weight for choosing a Guard+Exit for the guard position.
1653 Let Wmd be the weight for choosing a Guard+Exit for the middle position.
1654 Let Wed be the weight for choosing a Guard+Exit for the exit position.
1656 Let Wme be the weight for choosing an Exit for the middle position.
1657 Let Wmg be the weight for choosing a Guard for the middle position.
1659 Let Wgg be the weight for choosing a Guard for the guard position.
1660 Let Wee be the weight for choosing an Exit for the exit position.
1662 Balanced network conditions then arise from solutions to the following
1663 system of equations:
1665 Wgg*G + Wgd*D == M + Wmd*D + Wme*E + Wmg*G (guard bw = middle bw)
1666 Wgg*G + Wgd*D == Wee*E + Wed*D (guard bw = exit bw)
1667 Wed*D + Wmd*D + Wgd*D == D (aka: Wed+Wmd+Wdg = 1)
1668 Wmg*G + Wgg*G == G (aka: Wgg = 1-Wmg)
1669 Wme*E + Wee*E == E (aka: Wee = 1-Wme)
1671 We are short 2 constraints with the above set. The remaining constraints
1672 come from examining different cases of network load.
1674 Case 1: E >= T/3 && G >= T/3 (Neither Exit nor Guard Scarce)
1676 In this case, the additional two constraints are: Wme*E == Wmd*D and
1677 Wgd == 0, which maximizes Exit-flagged bandwidth in the middle position.
1679 This leads to the solution:
1681 Wgg = (weight_scale*(D+E+G+M))/(3*G)
1682 Wmd = (weight_scale*(2*D + 2*E - G - M))/(6*D)
1683 Wme = (weight_scale*(2*D + 2*E - G - M))/(6*E)
1684 Wee = (weight_scale*(-2*D + 4*E + G + M))/(6*E)
1685 Wmg = weight_scale - Wgg
1686 Wed = weight_scale - Wmd
1687 Wgd = 0
1689 Case 2: E < T/3 && G < T/3 (Both are scarce)
1691 Let R denote the more scarce class (Rare) between Guard vs Exit.
1692 Let S denote the less scarce class.
1694 Subcase a: R+D < S
1696 In this subcase, we simply devote all of D bandwidth to the
1697 scarce class.
1699 Wgg = Wee = weight_scale
1700 Wmg = Wme = Wmd = 0;
1701 if E < G:
1702 Wed = weight_scale
1703 Wgd = 0
1704 else:
1705 Wed = 0
1706 Wgd = weight_scale
1708 Subcase b: R+D >= S
1710 In this case, if M <= T/3, we have enough bandwidth to try to achieve
1711 a balancing condition, and add the constraints Wgg == 1 and
1712 Wme*E == Wmd*D:
1714 Wgg = weight_scale
1715 Wgd = (weight_scale*(D + E - 2*G + M))/(3*D) (T/3 >= G (Ok))
1716 Wmd = (weight_scale*(D + E + G - 2*M))/(6*D) (T/3 >= M)
1717 Wme = (weight_scale*(D + E + G - 2*M))/(6*E)
1718 Wee = (weight_scale*(-D + 5*E - G + 2*M))/(6*E) (2E+M >= T/3)
1719 Wmg = 0;
1720 Wed = weight_scale - Wgd - Wmd
1722 If M >= T/3, the above solution will not be valid (one of the weights
1723 will be < 0 or > 1). In this case, we use:
1725 Wgg = weight_scale
1726 Wee = weight_scale
1727 Wmg = Wme = Wmd = 0
1728 Wgd = (weight_scale*(D+E-G))/(2*D)
1729 Wed = weight_scale - Wgd
1731 Case 3: One of E < T/3 or G < T/3
1733 Let S be the scarce class (of E or G).
1735 Subcase a: (S+D) < T/3:
1736 if G=S:
1737 Wgg = Wgd = weight_scale;
1738 Wmd = Wed = Wmg = 0;
1739 Wme = (weight_scale*(E-M))/(2*E);
1740 Wee = weight_scale-Wme;
1741 if E=S:
1742 Wee = Wed = weight_scale;
1743 Wmd = Wgd = Wmg = 0;
1744 Wmg = (weight_scale*(G-M))/(2*G);
1745 Wgg = weight_scale-Wmg;
1747 Subcase b: (S+D) >= T/3
1748 if G=S:
1749 Add constraints Wmg = 0, Wme*E == Wmd*D to maximize exit bandwidth
1750 in the middle position:
1751 Wgd = (weight_scale*(D + E - 2*G + M))/(3*D);
1752 Wmd = (weight_scale*(D + E + G - 2*M))/(6*D);
1753 Wme = (weight_scale*(D + E + G - 2*M))/(6*E);
1754 Wee = (weight_scale*(-D + 5*E - G + 2*M))/(6*E);
1755 Wgg = weight_scale;
1756 Wmg = 0;
1757 Wed = weight_scale - Wgd - Wmd;
1758 if E=S:
1759 Add constraints Wgd = 0, Wme*E == Wmd*D:
1760 Wgg = (weight_scale*(D + E + G + M))/(3*G);
1761 Wmd = (weight_scale*(2*D + 2*E - G - M))/(6*D);
1762 Wme = (weight_scale*(2*D + 2*E - G - M))/(6*E);
1763 Wee = (weight_scale*(-2*D + 4*E + G + M))/(6*E);
1764 Wgd = 0;
1765 Wmg = weight_scale - Wgg;
1766 Wed = weight_scale - Wmd;
1768 To ensure consensus, all calculations are performed using integer math
1769 with a fixed precision determined by the bwweightscale consensus
1770 parameter (defaults at 10000).
1772 For future balancing improvements, Tor clients support 11 additional weights
1773 for directory requests and middle weighting. These weights are currently
1774 set at weight_scale, with the exception of the following groups of
1775 assignments:
1777 Directory requests use middle weights:
1778 Wbd=Wmd, Wbg=Wmg, Wbe=Wme, Wbm=Wmm
1780 Handle bridges and strange exit policies:
1781 Wgm=Wgg, Wem=Wee, Weg=Wed
1783 3.5. Detached signatures
1785 Assuming full connectivity, every authority should compute and sign the
1786 same consensus directory in each period. Therefore, it isn't necessary to
1787 download the consensus computed by each authority; instead, the
1788 authorities only push/fetch each others' signatures. A "detached
1789 signature" document contains items as follows:
1791 "consensus-digest" SP Digest NL
1793 [At start, at most once.]
1795 The digest of the consensus being signed.
1797 "valid-after" SP YYYY-MM-DD SP HH:MM:SS NL
1798 "fresh-until" SP YYYY-MM-DD SP HH:MM:SS NL
1799 "valid-until" SP YYYY-MM-DD SP HH:MM:SS NL
1801 [As in the consensus]
1803 "directory-signature"
1805 [As in the consensus; the signature object is the same as in the
1806 consensus document.]
1809 4. Directory server operation
1811 All directory authorities and directory caches ("directory servers")
1812 implement this section, except as noted.
1814 4.1. Accepting uploads (authorities only)
1816 When a router posts a signed descriptor to a directory authority, the
1817 authority first checks whether it is well-formed and correctly
1818 self-signed. If it is, the authority next verifies that the nickname
1819 in question is not already assigned to a router with a different
1820 public key.
1821 Finally, the authority MAY check that the router is not blacklisted
1822 because of its key, IP, or another reason.
1824 If the descriptor passes these tests, and the authority does not already
1825 have a descriptor for a router with this public key, it accepts the
1826 descriptor and remembers it.
1828 If the authority _does_ have a descriptor with the same public key, the
1829 newly uploaded descriptor is remembered if its publication time is more
1830 recent than the most recent old descriptor for that router, and either:
1831 - There are non-cosmetic differences between the old descriptor and the
1832 new one.
1833 - Enough time has passed between the descriptors' publication times.
1834 (Currently, 12 hours.)
1836 Differences between router descriptors are "non-cosmetic" if they would be
1837 sufficient to force an upload as described in section 2 above.
1839 Note that the "cosmetic difference" test only applies to uploaded
1840 descriptors, not to descriptors that the authority downloads from other
1841 authorities.
1843 When a router posts a signed extra-info document to a directory authority,
1844 the authority again checks it for well-formedness and correct signature,
1845 and checks that its matches the extra-info-digest in some router
1846 descriptor that it believes is currently useful. If so, it accepts it and
1847 stores it and serves it as requested. If not, it drops it.
1849 4.2. Voting (authorities only)
1851 Authorities divide time into Intervals. Authority administrators SHOULD
1852 try to all pick the same interval length, and SHOULD pick intervals that
1853 are commonly used divisions of time (e.g., 5 minutes, 15 minutes, 30
1854 minutes, 60 minutes, 90 minutes). Voting intervals SHOULD be chosen to
1855 divide evenly into a 24-hour day.
1857 Authorities SHOULD act according to interval and delays in the
1858 latest consensus. Lacking a latest consensus, they SHOULD default to a
1859 30-minute Interval, a 5 minute VotingDelay, and a 5 minute DistDelay.
1861 Authorities MUST take pains to ensure that their clocks remain accurate
1862 within a few seconds. (Running NTP is usually sufficient.)
1864 The first voting period of each day begins at 00:00 (midnight) GMT. If
1865 the last period of the day would be truncated by one-half or more, it is
1866 merged with the second-to-last period.
1868 An authority SHOULD publish its vote immediately at the start of each voting
1869 period (minus VoteSeconds+DistSeconds). It does this by making it
1870 available at
1871 http://<hostname>/tor/status-vote/next/authority.z
1872 and sending it in an HTTP POST request to each other authority at the URL
1873 http://<hostname>/tor/post/vote
1875 If, at the start of the voting period, minus DistSeconds, an authority
1876 does not have a current statement from another authority, the first
1877 authority downloads the other's statement.
1879 Once an authority has a vote from another authority, it makes it available
1880 at
1881 http://<hostname>/tor/status-vote/next/<fp>.z
1882 where <fp> is the fingerprint of the other authority's identity key.
1883 And at
1884 http://<hostname>/tor/status-vote/next/d/<d>.z
1885 where <d> is the digest of the vote document.
1887 The consensus status, along with as many signatures as the server
1888 currently knows, should be available at
1889 http://<hostname>/tor/status-vote/next/consensus.z
1890 All of the detached signatures it knows for consensus status should be
1891 available at:
1892 http://<hostname>/tor/status-vote/next/consensus-signatures.z
1894 Once there are enough signatures, or once the voting period starts,
1895 these documents are available at
1896 http://<hostname>/tor/status-vote/current/consensus.z
1897 and
1898 http://<hostname>/tor/status-vote/current/consensus-signatures.z
1899 [XXX current/consensus-signatures is not currently implemented, as it
1900 is not used in the voting protocol.]
1902 The other vote documents are analogously made available under
1903 http://<hostname>/tor/status-vote/current/authority.z
1904 http://<hostname>/tor/status-vote/current/<fp>.z
1905 http://<hostname>/tor/status-vote/current/d/<d>.z
1906 once the consensus is complete.
1908 Once an authority has computed and signed a consensus network status, it
1909 should send its detached signature to each other authority in an HTTP POST
1910 request to the URL:
1911 http://<hostname>/tor/post/consensus-signature
1913 [XXX Note why we support push-and-then-pull.]
1915 [XXX possible future features include support for downloading old
1916 consensuses.]
1918 4.3. Downloading consensus status documents (caches only)
1920 All directory servers (authorities and caches) try to keep a recent
1921 network-status consensus document to serve to clients. A cache ALWAYS
1922 downloads a network-status consensus if any of the following are true:
1923 - The cache has no consensus document.
1924 - The cache's consensus document is no longer valid.
1925 Otherwise, the cache downloads a new consensus document at a randomly
1926 chosen time in the first half-interval after its current consensus
1927 stops being fresh. (This time is chosen at random to avoid swarming
1928 the authorities at the start of each period. The interval size is
1929 inferred from the difference between the valid-after time and the
1930 fresh-until time on the consensus.)
1932 [For example, if a cache has a consensus that became valid at 1:00,
1933 and is fresh until 2:00, that cache will fetch a new consensus at
1934 a random time between 2:00 and 2:30.]
1936 4.4. Downloading and storing router descriptors (authorities and caches)
1938 Periodically (currently, every 10 seconds), directory servers check
1939 whether there are any specific descriptors that they do not have and that
1940 they are not currently trying to download. Caches identify these
1941 descriptors by hash in the recent network-status consensus documents;
1942 authorities identify them by hash in vote (if publication date is more
1943 recent than the descriptor we currently have).
1945 [XXXX need a way to fetch descriptors ahead of the vote? v2 status docs can
1946 do that for now.]
1948 If so, the directory server launches requests to the authorities for these
1949 descriptors, such that each authority is only asked for descriptors listed
1950 in its most recent vote (if the requester is an authority) or in the
1951 consensus (if the requester is a cache). If we're an authority, and more
1952 than one authority lists the descriptor, we choose which to ask at random.
1954 If one of these downloads fails, we do not try to download that descriptor
1955 from the authority that failed to serve it again unless we receive a newer
1956 network-status (consensus or vote) from that authority that lists the same
1957 descriptor.
1959 Directory servers must potentially cache multiple descriptors for each
1960 router. Servers must not discard any descriptor listed by any recent
1961 consensus. If there is enough space to store additional descriptors,
1962 servers SHOULD try to hold those which clients are likely to download the
1963 most. (Currently, this is judged based on the interval for which each
1964 descriptor seemed newest.)
1965 [XXXX define recent]
1967 Authorities SHOULD NOT download descriptors for routers that they would
1968 immediately reject for reasons listed in 3.1.
1970 4.5. Downloading and storing extra-info documents
1972 All authorities, and any cache that chooses to cache extra-info documents,
1973 and any client that uses extra-info documents, should implement this
1974 section.
1976 Note that generally, clients don't need extra-info documents.
1978 Periodically, the Tor instance checks whether it is missing any extra-info
1979 documents: in other words, if it has any router descriptors with an
1980 extra-info-digest field that does not match any of the extra-info
1981 documents currently held. If so, it downloads whatever extra-info
1982 documents are missing. Caches download from authorities; non-caches try
1983 to download from caches. We follow the same splitting and back-off rules
1984 as in 4.4 (if a cache) or 5.3 (if a client).
1986 4.6. General-use HTTP URLs
1988 "Fingerprints" in these URLs are base-16-encoded SHA1 hashes.
1990 The most recent v3 consensus should be available at:
1991 http://<hostname>/tor/status-vote/current/consensus.z
1993 Starting with Tor version 0.2.1.1-alpha is also available at:
1994 http://<hostname>/tor/status-vote/current/consensus/<F1>+<F2>+<F3>.z
1996 Where F1, F2, etc. are authority identity fingerprints the client trusts.
1997 Servers will only return a consensus if more than half of the requested
1998 authorities have signed the document, otherwise a 404 error will be sent
1999 back. The fingerprints can be shortened to a length of any multiple of
2000 two, using only the leftmost part of the encoded fingerprint. Tor uses
2001 3 bytes (6 hex characters) of the fingerprint.
2003 Clients SHOULD sort the fingerprints in ascending order. Server MUST
2004 accept any order.
2006 Clients SHOULD use this format when requesting consensus documents from
2007 directory authority servers and from caches running a version of Tor
2008 that is known to support this URL format.
2010 A concatenated set of all the current key certificates should be available
2011 at:
2012 http://<hostname>/tor/keys/all.z
2014 The key certificate for this server (if it is an authority) should be
2015 available at:
2016 http://<hostname>/tor/keys/authority.z
2018 The key certificate for an authority whose authority identity fingerprint
2019 is <F> should be available at:
2020 http://<hostname>/tor/keys/fp/<F>.z
2022 The key certificate whose signing key fingerprint is <F> should be
2023 available at:
2024 http://<hostname>/tor/keys/sk/<F>.z
2026 The key certificate whose identity key fingerprint is <F> and whose signing
2027 key fingerprint is <S> should be available at:
2029 http://<hostname>/tor/keys/fp-sk/<F>-<S>.z
2031 (As usual, clients may request multiple certificates using:
2032 http://<hostname>/tor/keys/fp-sk/<F1>-<S1>+<F2>-<S2>.z )
2033 [The above fp-sk format was not supported before Tor 0.2.1.9-alpha.]
2035 The most recent descriptor for a server whose identity key has a
2036 fingerprint of <F> should be available at:
2037 http://<hostname>/tor/server/fp/<F>.z
2039 The most recent descriptors for servers with identity fingerprints
2040 <F1>,<F2>,<F3> should be available at:
2041 http://<hostname>/tor/server/fp/<F1>+<F2>+<F3>.z
2043 (NOTE: Implementations SHOULD NOT download descriptors by identity key
2044 fingerprint. This allows a corrupted server (in collusion with a cache) to
2045 provide a unique descriptor to a client, and thereby partition that client
2046 from the rest of the network.)
2048 The server descriptor with (descriptor) digest <D> (in hex) should be
2049 available at:
2050 http://<hostname>/tor/server/d/<D>.z
2052 The most recent descriptors with digests <D1>,<D2>,<D3> should be
2053 available at:
2054 http://<hostname>/tor/server/d/<D1>+<D2>+<D3>.z
2056 The most recent descriptor for this server should be at:
2057 http://<hostname>/tor/server/authority.z
2058 [Nothing in the Tor protocol uses this resource yet, but it is useful
2059 for debugging purposes. Also, the official Tor implementations
2060 (starting at 0.1.1.x) use this resource to test whether a server's
2061 own DirPort is reachable.]
2063 A concatenated set of the most recent descriptors for all known servers
2064 should be available at:
2065 http://<hostname>/tor/server/all.z
2067 Extra-info documents are available at the URLS
2068 http://<hostname>/tor/extra/d/...
2069 http://<hostname>/tor/extra/fp/...
2070 http://<hostname>/tor/extra/all[.z]
2071 http://<hostname>/tor/extra/authority[.z]
2072 (As for /tor/server/ URLs: supports fetching extra-info
2073 documents by their digest, by the fingerprint of their servers,
2074 or all at once. When serving by fingerprint, we serve the
2075 extra-info that corresponds to the descriptor we would serve by
2076 that fingerprint. Only directory authorities of version
2077 0.2.0.1-alpha or later are guaranteed to support the first
2078 three classes of URLs. Caches may support them, and MUST
2079 support them if they have advertised "caches-extra-info".)
2081 For debugging, directories SHOULD expose non-compressed objects at URLs like
2082 the above, but without the final ".z".
2083 Clients MUST handle compressed concatenated information in two forms:
2084 - A concatenated list of zlib-compressed objects.
2085 - A zlib-compressed concatenated list of objects.
2086 Directory servers MAY generate either format: the former requires less
2087 CPU, but the latter requires less bandwidth.
2089 Clients SHOULD use upper case letters (A-F) when base16-encoding
2090 fingerprints. Servers MUST accept both upper and lower case fingerprints
2091 in requests.
2093 5. Client operation: downloading information
2095 Every Tor that is not a directory server (that is, those that do
2096 not have a DirPort set) implements this section.
2098 5.1. Downloading network-status documents
2100 Each client maintains a list of directory authorities. Insofar as
2101 possible, clients SHOULD all use the same list.
2103 Clients try to have a live consensus network-status document at all times.
2104 A network-status document is "live" if the time in its valid-until field
2105 has not passed.
2107 If a client is missing a live network-status document, it tries to fetch
2108 it from a directory cache (or from an authority if it knows no caches).
2109 On failure, the client waits briefly, then tries that network-status
2110 document again from another cache. The client does not build circuits
2111 until it has a live network-status consensus document, and it has
2112 descriptors for more than 1/4 of the routers that it believes are running.
2114 (Note: clients can and should pick caches based on the network-status
2115 information they have: once they have first fetched network-status info
2116 from an authority, they should not need to go to the authority directly
2117 again.)
2119 To avoid swarming the caches whenever a consensus expires, the
2120 clients download new consensuses at a randomly chosen time after the
2121 caches are expected to have a fresh consensus, but before their
2122 consensus will expire. (This time is chosen uniformly at random from
2123 the interval between the time 3/4 into the first interval after the
2124 consensus is no longer fresh, and 7/8 of the time remaining after
2125 that before the consensus is invalid.)
2127 [For example, if a cache has a consensus that became valid at 1:00,
2128 and is fresh until 2:00, and expires at 4:00, that cache will fetch
2129 a new consensus at a random time between 2:45 and 3:50, since 3/4
2130 of the one-hour interval is 45 minutes, and 7/8 of the remaining 75
2131 minutes is 65 minutes.]
2133 5.2. Downloading and storing router descriptors
2135 Clients try to have the best descriptor for each router. A descriptor is
2136 "best" if:
2137 * It is listed in the consensus network-status document.
2139 Periodically (currently every 10 seconds) clients check whether there are
2140 any "downloadable" descriptors. A descriptor is downloadable if:
2141 - It is the "best" descriptor for some router.
2142 - The descriptor was published at least 10 minutes in the past.
2143 (This prevents clients from trying to fetch descriptors that the
2144 mirrors have probably not yet retrieved and cached.)
2145 - The client does not currently have it.
2146 - The client is not currently trying to download it.
2147 - The client would not discard it immediately upon receiving it.
2148 - The client thinks it is running and valid (see 6.1 below).
2150 If at least 16 known routers have downloadable descriptors, or if
2151 enough time (currently 10 minutes) has passed since the last time the
2152 client tried to download descriptors, it launches requests for all
2153 downloadable descriptors, as described in 5.3 below.
2155 When a descriptor download fails, the client notes it, and does not
2156 consider the descriptor downloadable again until a certain amount of time
2157 has passed. (Currently 0 seconds for the first failure, 60 seconds for the
2158 second, 5 minutes for the third, 10 minutes for the fourth, and 1 day
2159 thereafter.) Periodically (currently once an hour) clients reset the
2160 failure count.
2162 Clients retain the most recent descriptor they have downloaded for each
2163 router so long as it is not too old (currently, 48 hours), OR so long as
2164 no better descriptor has been downloaded for the same router.
2166 [Versions of Tor before 0.1.2.3-alpha would discard descriptors simply for
2167 being published too far in the past.] [The code seems to discard
2168 descriptors in all cases after they're 5 days old. True? -RD]
2170 5.3. Managing downloads
2172 When a client has no consensus network-status document, it downloads it
2173 from a randomly chosen authority. In all other cases, the client
2174 downloads from caches randomly chosen from among those believed to be V2
2175 directory servers. (This information comes from the network-status
2176 documents; see 6 below.)
2178 When downloading multiple router descriptors, the client chooses multiple
2179 mirrors so that:
2180 - At least 3 different mirrors are used, except when this would result
2181 in more than one request for under 4 descriptors.
2182 - No more than 128 descriptors are requested from a single mirror.
2183 - Otherwise, as few mirrors as possible are used.
2184 After choosing mirrors, the client divides the descriptors among them
2185 randomly.
2187 After receiving any response client MUST discard any network-status
2188 documents and descriptors that it did not request.
2190 6. Using directory information
2192 Everyone besides directory authorities uses the approaches in this section
2193 to decide which servers to use and what their keys are likely to be.
2194 (Directory authorities just believe their own opinions, as in 3.1 above.)
2196 6.1. Choosing routers for circuits.
2198 Circuits SHOULD NOT be built until the client has enough directory
2199 information: a live consensus network status [XXXX fallback?] and
2200 descriptors for at least 1/4 of the servers believed to be running.
2202 A server is "listed" if it is included by the consensus network-status
2203 document. Clients SHOULD NOT use unlisted servers.
2205 These flags are used as follows:
2207 - Clients SHOULD NOT use non-'Valid' or non-'Running' routers unless
2208 requested to do so.
2210 - Clients SHOULD NOT use non-'Fast' routers for any purpose other than
2211 very-low-bandwidth circuits (such as introduction circuits).
2213 - Clients SHOULD NOT use non-'Stable' routers for circuits that are
2214 likely to need to be open for a very long time (such as those used for
2215 IRC or SSH connections).
2217 - Clients SHOULD NOT choose non-'Guard' nodes when picking entry guard
2218 nodes.
2220 - Clients SHOULD NOT download directory information from non-'V2Dir'
2221 caches.
2223 See the "path-spec.txt" document for more details.
2225 6.2. Managing naming
2227 In order to provide human-memorable names for individual server
2228 identities, some directory servers bind names to IDs. Clients handle
2229 names in two ways:
2231 When a client encounters a name it has not mapped before:
2233 If the consensus lists any router with that name as "Named", or if
2234 consensus-method 2 or later is in use and the consensus lists any
2235 router with that name as having the "Unnamed" flag, then the name is
2236 bound. (It's bound to the ID listed in the entry with the Named,
2237 or to an unknown ID if no name is found.)
2239 When the user refers to a bound name, the implementation SHOULD provide
2240 only the router with ID bound to that name, and no other router, even
2241 if the router with the right ID can't be found.
2243 When a user tries to refer to a non-bound name, the implementation SHOULD
2244 warn the user. After warning the user, the implementation MAY use any
2245 router that advertises the name.
2247 Not every router needs a nickname. When a router doesn't configure a
2248 nickname, it publishes with the default nickname "Unnamed". Authorities
2249 SHOULD NOT ever mark a router with this nickname as Named; client software
2250 SHOULD NOT ever use a router in response to a user request for a router
2251 called "Unnamed".
2253 6.3. Software versions
2255 An implementation of Tor SHOULD warn when it has fetched a consensus
2256 network-status, and it is running a software version not listed.
2258 6.4. Warning about a router's status.
2260 If a router tries to publish its descriptor to a Naming authority
2261 that has its nickname mapped to another key, the router SHOULD
2262 warn the operator that it is either using the wrong key or is using
2263 an already claimed nickname.
2265 If a router has fetched a consensus document,, and the
2266 authorities do not publish a binding for the router's nickname, the
2267 router MAY remind the operator that the chosen nickname is not
2268 bound to this key at the authorities, and suggest contacting the
2269 authority operators.
2271 ...
2273 6.5. Router protocol versions
2275 A client should believe that a router supports a given feature if that
2276 feature is supported by the router or protocol versions in more than half
2277 of the live networkstatuses' "v" entries for that router. In other words,
2278 if the "v" entries for some router are:
2279 v Tor 0.0.8pre1 (from authority 1)
2280 v Tor 0.1.2.11 (from authority 2)
2281 v FutureProtocolDescription 99 (from authority 3)
2282 then the client should believe that the router supports any feature
2283 supported by 0.1.2.11.
2285 This is currently equivalent to believing the median declared version for
2286 a router in all live networkstatuses.
2288 7. Standards compliance
2290 All clients and servers MUST support HTTP 1.0. Clients and servers MAY
2291 support later versions of HTTP as well.
2293 7.1. HTTP headers
2295 Servers MAY set the Content-Length: header. Servers SHOULD set
2296 Content-Encoding to "deflate" or "identity".
2298 Servers MAY include an X-Your-Address-Is: header, whose value is the
2299 apparent IP address of the client connecting to them (as a dotted quad).
2300 For directory connections tunneled over a BEGIN_DIR stream, servers SHOULD
2301 report the IP from which the circuit carrying the BEGIN_DIR stream reached
2302 them. [Servers before version 0.1.2.5-alpha reported 127.0.0.1 for all
2303 BEGIN_DIR-tunneled connections.]
2305 Servers SHOULD disable caching of multiple network statuses or multiple
2306 router descriptors. Servers MAY enable caching of single descriptors,
2307 single network statuses, the list of all router descriptors, a v1
2308 directory, or a v1 running routers document. XXX mention times.
2310 7.2. HTTP status codes
2312 Tor delivers the following status codes. Some were chosen without much
2313 thought; other code SHOULD NOT rely on specific status codes yet.
2315 200 -- the operation completed successfully
2316 -- the user requested statuses or serverdescs, and none of the ones we
2317 requested were found (0.2.0.4-alpha and earlier).
2319 304 -- the client specified an if-modified-since time, and none of the
2320 requested resources have changed since that time.
2322 400 -- the request is malformed, or
2323 -- the URL is for a malformed variation of one of the URLs we support,
2324 or
2325 -- the client tried to post to a non-authority, or
2326 -- the authority rejected a malformed posted document, or
2328 404 -- the requested document was not found.
2329 -- the user requested statuses or serverdescs, and none of the ones
2330 requested were found (0.2.0.5-alpha and later).
2332 503 -- we are declining the request in order to save bandwidth
2333 -- user requested some items that we ordinarily generate or store,
2334 but we do not have any available.
2336 9. Backward compatibility and migration plans
2338 Until Tor versions before 0.1.1.x are completely obsolete, directory
2339 authorities should generate, and mirrors should download and cache, v1
2340 directories and running-routers lists, and allow old clients to download
2341 them. These documents and the rules for retrieving, serving, and caching
2342 them are described in dir-spec-v1.txt.
2344 Until Tor versions before 0.2.0.x are completely obsolete, directory
2345 authorities should generate, mirrors should download and cache, v2
2346 network-status documents, and allow old clients to download them.
2347 Additionally, all directory servers and caches should download, store, and
2348 serve any router descriptor that is required because of v2 network-status
2349 documents. These documents and the rules for retrieving, serving, and
2350 caching them are described in dir-spec-v1.txt.
2352 A. Consensus-negotiation timeline.
2354 Period begins: this is the Published time.
2355 Everybody sends votes
2356 Reconciliation: everybody tries to fetch missing votes.
2357 consensus may exist at this point.
2358 End of voting period:
2359 everyone swaps signatures.
2360 Now it's okay for caches to download
2361 Now it's okay for clients to download.
2363 Valid-after/valid-until switchover