tor-network-consensus-guards-and-paths
Use when reasoning about the Tor network as an operated system — relays and flags, the nine directory authorities, the hourly consensus, bandwidth weights, guard pinning and why it exists, path-selection restrictions, and happy families — or when looking up a relay with Relay Search, Onionoo, ExoneraTor or CollecTor. Covers §6–§7. Companion to the other Tor and onion-network skills.
- License 许可
- MIT
- 最近更新
- 2026-11-15
- Stars
- ★ 5
- Forks
- 3
01 前言 Front Matter
- name
- tor-network-consensus-guards-and-paths
- description
- Use when reasoning about the Tor network as an operated system — relays and flags, the nine directory authorities, the hourly consensus, bandwidth weights, guard pinning and why it exists, path-selection restrictions, and happy families — or when looking up a relay with Relay Search, Onionoo, ExoneraTor or CollecTor. Covers §6–§7. Companion to the other Tor and onion-network skills.
03 SKILL.md 正文
The Tor Network: Relays, Authorities, Consensus, and Path Selection
Part 2 of 8 of the Tor and Onion Networks reference (plugin
tor-and-onion-networks), covering §6–§7. Sibling skills:tor-protocol-circuits-and-cell-cryptography(§1–§5),tor-onion-services-and-hardening(§8–§9),tor-censorship-circumvention-and-bridges(§10),tor-attack-literature-and-threat-model(§11),tor-ecosystem-alternatives-and-governance(§12, §14–§16),tor-building-on-tor-software-patterns(§13.1–§13.4, handbook §1),tor-running-relays-bridges-and-hardware(§13.5, handbook §2–§6). Section numbers are shared across the set; a reference written as §N →skillpoints into that sibling skill.Currency: compiled 16 September 2026. Guard logic and consensus mechanics are stable. Relay counts, user counts and version-gated features are dated — pull live figures from Tor Metrics rather than quoting this document.
⚠️ Tor's trust root is nine servers run by nine named people. Everything else is volunteers.
⚠️ GOTCHA boxes mark where the mental model people carry is wrong in ways that cause real deanonymization or bad operational decisions.
The three ideas that organize this document:
- ⚠️ THE CONSENSUS IS THE NETWORK (§6.2). There is no peer-to-peer discovery. Nine directory authorities vote hourly on who exists and how much they weigh, and clients trust that signed document. Majority compromise is the nightmare scenario, which is why authority diversity is a governance question, not an engineering one.
- ⚠️ GUARDS TRADE A CERTAINTY FOR A RARITY (§7.1). Rotating first hops randomly means an adversary with a modest relay fraction eventually becomes your entry. Pinning a small guard set for months converts "eventually certain" into "probably never" — and is why New Identity deliberately does not change your guard.
- ⚠️ PATH RESTRICTIONS ARE ANTI-CORRELATION, NOT PRIVACY CONTROLS (§7.2). Family and /16 rules exist so one operator cannot own two hops.
ExitNodescountry pinning is advisory routing preference, not a safety property — the exit still sees your traffic.
§6. The network itself: relays, authorities, consensus, weights
§6.1 Actors
- Relays (~9,200 running in mid-September 2026; Tor Metrics
networksize.csv: 10,065 on 20 Aug 2026 declining to 9,228 by 14 Sept — treat single weeks as noisy): volunteer-operated nodes. Flags matter:Guard(stable, fast, trustworthy enough to be first hops),Exit(allowed final hops),HSDir(hold onion-service descriptors),Fast,Stable,Running,Valid,BadExit. - Directory authorities (9). The trusted core: nine long-lived servers run by different known individuals/organizations (moria1, tor26, dizum, gabelmoo, dannenberg, maatuska, Faravahar, longclaw, bastet). Every hour they vote on the set of relays and compute a consensus — the signed, authoritative network document. Five-of-nine agreement is what keeps the network self-consistent; compromise of the majority is the nightmare scenario, and it's why authority diversity and physical security are taken so seriously.
- Fallback directory mirrors (~100+ hard-coded relays) so new clients can fetch a consensus even if they've never heard of an authority.
- Bandwidth authorities. A subset of authorities additionally run sbws scanners that actively measure relay throughput; their votes set consensus weights so clients pick relays in proportion to real capacity.
- Bridges (~2,300 in Sept 2026): unlisted guard-like entry nodes for censored users,
distributed out-of-band (§10 →
tor-censorship-circumvention-and-bridges).
§6.2 Consensus mechanics
Hourly cycle: authorities publish votes (~:50 past the hour), exchange them, and produce a consensus valid for three hours. Clients and relays download microdescriptors (compact relay records keyed by digest) rather than full descriptors — this is what made Tor usable on mobile links. Consensus methods version the format itself; 0.4.9 dropped support for consensus methods older than 32. A consensus transparency experiment (signed-tree logging, CT-style, to detect split-view attacks against directory clients) was underway as of the 0.4.9 series, with unsigned-consensus export support added for it.
§6.3 Users
Tor Metrics estimates (from directory-request extrapolation, requests/10 per client per day —
see their reproducible metrics
methodology): ~2.8–3.5 million
directly-connecting users per day in September 2026, plus bridge users counted separately.
Largest populations (Sept 2026 top-10 table): US (~515k, ~15.5%), Germany (~294k), Brazil,
Finland, India, France. Usage spikes historically track censorship events and conflicts (Iran
June 2025 being the sharpest recent example — §10.4 →
tor-censorship-circumvention-and-bridges).
⚠️ GOTCHA: "Tor user counts" are estimates derived from directory requests, not logins. They are extrapolations with a documented methodology and known error bars. Never present them as a census, and never compare across a methodology change without reading the metrics notes.
§6.4 Reading the network yourself
- Relay Search — per-relay detail: flags, weights, family, history.
- Onionoo — the REST/JSON API behind Relay Search and most third-party tooling.
- CollecTor — descriptor archives for research.
- ExoneraTor — "was IP X a Tor exit on date Y?" (the answer to most "was it Tor?" abuse questions).
- Exit list / DNSBL — live exit enumeration for destination-side filtering design.
§7. Guard selection and path restrictions
§7.1 Why guards exist
The 2013 paper Users Get Routed (Johnson et al., CCS 2013) quantified what everyone feared: if clients rotate first hops randomly, an adversary running a modest fraction of relays eventually becomes the first hop and can run confirmation attacks. The fix — pin a small set of first hops for months — turned a probabilistic certainty into a probabilistic rarity.
Modern rules (proposal 271 and successors): a sampled guard set, filtered for reachability, of which the client uses 2–3 primary guards, rotating only on failure or expiry.
⚠️ GOTCHA: "Get a new identity often" is not the same as changing guards, and Tor Browser's New Identity keeps your guards deliberately. Advice that tells users to churn entry nodes for safety inverts the actual security argument — churn is what guards exist to prevent.
§7.2 Path selection restrictions
Circuits are built subject to constraints that any builder must replicate:
- relays in one circuit must not share a family or (by default) a /16;
- exit policies must permit the destination port;
- consensus weights bias selection toward capacity (and
Guard/Exitfractional weights keep exit-scarce bandwidth available for exit use); - country restrictions only via explicit torrc (
ExitNodes,ExcludeNodes,GeoIP...) — and these are advisory for privacy, not robust safety properties, since the exit or a router downstream sees your traffic anyway.
§7.3 Happy families (proposal 321; shipped in C tor 0.4.9, 2026)
MyFamily mutual cross-listing was error-prone and bloated descriptors (O(family²)
fingerprints). 0.4.9 adds a shared family key: relays sign a certificate proving family
membership, collapsing family proof to a single identifier. Tor Project estimates eventual
~80% microdescriptor size reduction.
⚠️ GOTCHA for relay operators: until client adoption catches up, you must maintain
MyFamilyin parallel with the new family key. DroppingMyFamilyearly makes your relays look unrelated to older clients, which is exactly the correlation risk families exist to prevent. See the family-ids documentation.
Where to go next
- How cells and circuits actually work underneath all of this: §3 →
tor-protocol-circuits-and-cell-cryptography. - Why onion services need extra guard layers on top of this: §9.1 →
tor-onion-services-and-hardening. - Sybil campaigns and malicious relays against this trust model: §11.4 →
tor-attack-literature-and-threat-model. - Running a relay, bridge or exit yourself: §13.5 →
tor-running-relays-bridges-and-hardware.
04 同仓库更多 Skills
white-paper-writing
Use when writing white papers, technical reports, policy documents, thought leadership pieces, or any long-form authoritative document. Triggers on requests to write, structure, outline, or review white papers, technical reports, position papers, policy briefs, or evidence-driven arguments. Also use for B2B content marketing strategy, technical documentation requiring authority, academic-style reports, or any document that must educate while persuading. Use when someone asks "how should I structure this" for a long-form document.
prose-technical-mechanics
Use when editing or writing prose at a technical/mechanical level: choosing between Anglo-Saxon and Latinate diction, placing information in topic vs stress positions, writing periodic vs cumulative sentences, handling sentence rhythm, punctuation as logic, transitions, paragraph structure, or phonesthetics. Use when someone wants to understand WHY prose works or doesn't, not just surface-level style advice. Triggers on word choice, sentence structure, emphasis, rhythm, diction, or any deep writing mechanics question.
prose-mechanics
Use when writing, editing, or analyzing prose style; when trying to avoid generic AI-sounding writing; when applying specific prose styles such as journalistic, academic, narrative, KJV biblical, Hemingway plain, Latinate/Jamesian, Orwellian, Augustan, Victorian, Early Modern (Bacon, Browne, Milton), or free indirect discourse; or when coaching someone on writing voice and mechanics. Triggers on any request to improve, analyze, or write in a specific prose style, to fix "AI-sounding" writing, to make prose more vivid or concrete, to vary sentence rhythm, or to understand how a particular writer achieves their effects.
narrative-structure
Use when structuring stories, pitches, presentations, white papers, or any persuasive narrative; analyzing the hero's journey in fiction or non-fiction; applying StoryBrand for business communication; designing narrative arcs for thought leadership; or understanding Vogler, Snyder, Propp, or Campbell frameworks. Triggers on any narrative structure, storytelling, pitch structure, StoryBrand, hero's journey, three-act structure, Freytag, Save the Cat, narrative arc, or "how should I structure this story/pitch/presentation" request.
legal-brief-writing
Use when writing legal briefs, motions, or any court filing; when analyzing judicial cognition and persuasion; when applying elite brief-writing techniques from Roberts, Clement, Kagan, Garner; or when structuring legal arguments for maximum persuasion. Also use for autistic/neurodivergent advocates who need to understand how their natural communication style maps to elite legal advocacy. Triggers on any legal writing, brief drafting, motion preparation, or legal argument structuring request.
journal-publishability-criteria
Use whenever a research paper, article or preprint is being written, revised, reviewed or published, and specifically before every publication of docs/paper.md. Covers what legitimate journals actually gate on, in order of gating — the editor's desk screen (scope, novelty, package problems), what reviewers judge (methodological soundness, claims the evidence can carry, statistical and qualitative reporting), EQUATOR reporting guidelines, the ICMJE authorship criteria, COPE's view of text recycling, the near-universal generative-AI policy (no AI author, disclosure, author responsibility), TOP 2025 transparency standards, how to verify a journal is legitimate, submitting without an institution, and a pre-submission checklist. Triggers on requests to evaluate, review, prepare or submit a paper, to judge publishability, to write a declarations or data-availability section, to disclose AI use, or to check a journal.