312-relay-auto-ipv6-addr.txt

Filename: 312-relay-auto-ipv6-addr.txt
Title: Tor Relay Automatic IPv6 Address Discovery
Author: teor, Nick Mathewson, s7r
Created: 28-January-2020
Status: Accepted
Ticket: #33073

0. Abstract

   We propose that Tor relays (and bridges) should automatically find their
   IPv6 address.

   Like tor's existing IPv4 address auto-detection, the chosen IPv6 address
   will be published as an IPv6 ORPort in the relay's descriptor. Clients,
   relays, and authorities connect to relay descriptor IP addresses.
   Therefore, IP addresses in descriptors need to be publicly routable. (If
   the relay is running on the public tor network.)

   To discover their IPv6 address, some relays may fetch directory documents
   over IPv6. (For anonymity reasons, bridges are unable to fetch directory
   documents over IPv6, until clients start to do so.)

1. Introduction

   Tor relays (and bridges) currently find their IPv4 address, and use it as
   their ORPort and DirPort address when publishing their descriptor. But
   relays and bridges do not automatically find their IPv6 address.

   However, relay operators can manually configure an ORPort with an IPv6
   address, and that ORPort is published in their descriptor in an "or-address"
   line (see [Tor Directory Protocol]).

   Many relay operators don't know their relay's IPv4 or IPv6 addresses. So
   they rely on Tor's IPv4 auto-detection, and don't configure an IPv6
   address. When operators do configure an IPv6 address, it's easy for them to
   make mistakes. IPv6 ORPort issues are a significant source of relay
   operator support requests.

   Implementing IPv6 address auto-detection, and IPv6 ORPort reachability
   checks (see [Proposal 311: Relay IPv6 Reachability]) will increase the
   number of working IPv6-capable relays in the tor network.

2. Scope

   This proposal modifies Tor's behaviour as follows:

   Relays, bridges, and directory authorities:
     * automatically find their IPv6 address, and
     * for consistency between IPv4 and IPv6 detection:
       * start using IPv4 ORPort for IPv4 address detection, and
       * re-order IPv4 address detection methods.

   Relays (but not bridges, or directory authorities):
     * fetch some directory documents over IPv6.

   For anonymity reasons, bridges are unable to fetch directory documents over
   IPv6, until clients start to do so. (See
   [Proposal 306: Client Auto IPv6 Connections].)

   For security reasons, directory authorities must only use addresses that
   are explicitly configured in their torrc.

   This proposal makes a small, optional change to existing client behaviour:
     * clients also check IPv6 addresses when rotating TLS keys for new
       networks.
   In addition to the changes to IPv4 address resolution, most of which won't
   affect clients. (Because they do not set Address or ORPort.)

   Throughout this proposal, "relays" includes directory authorities, except
   where they are specifically excluded. "relays" does not include bridges,
   except where they are specifically included. (The first mention of "relays"
   in each section should specifically exclude or include these other roles.)

   When this proposal describes Tor's current behaviour, it covers all
   supported Tor versions (0.3.5.7 to 0.4.2.5), as of January 2020, except
   where another version is specifically mentioned.

3. Finding Relay IPv6 Addresses

   We propose that Tor relays (and bridges) should automatically find their
   IPv6 address.

   Like tor's existing IPv4 address auto-detection, the chosen IPv6 address
   will be published as an IPv6 ORPort in the relay's descriptor. Clients,
   relays, and authorities connect to relay descriptor IP addresses.
   Therefore, IP addresses in descriptors need to be publicly routable. (If
   the relay is running on the public tor network.)

   Relays should ignore any addresses that are reserved for private networks,
   and check the reachability of addresses that appear to be public (see
   [Proposal 311: Relay IPv6 Reachability]). Relays should only publish IP
   addresses in their descriptor, if they are public and reachable. (If the
   relay is not running on the public tor network, it may use any IP address.)

   To discover their IPv6 address, some relays may fetch directory documents
   over IPv6. (For anonymity reasons, bridges are unable to fetch directory
   documents over IPv6, until clients start to do so. For security reasons,
   directory authorities only use addresses that are explicitly configured in
   their torrc.)

3.1. Current Relay IPv4 Address Discovery

   Currently, all relays (and bridges) must have an IPv4 address. IPv6
   addresses are optional for relays.

   Tor currently tries to find relay IPv4 addresses in this order:
     1. the Address torrc option
     2. the address of the hostname (resolved using DNS, if needed)
     3. a local interface address
        (by making an unused socket, if needed)
     4. an address reported by a directory server (using X-Your-Address-Is)

   When using the Address option, or the hostname, tor supports:
     * an IPv4 address literal, or
     * resolving an IPv4 address from a hostname.

   If tor is running on the public network, and an address isn't globally
   routable, tor ignores it. (If it was explicitly set in Address, tor logs an
   error.)

   If there are multiple valid addresses, tor chooses:
     * the first address returned by the resolver,
     * the first address returned by the local interface API, and
     * the latest address(es) returned by a directory server, DNS, or the
       local interface API.

3.1.1. Current Relay IPv4 and IPv6 Address State Management

   Currently, relays (and bridges) manage their IPv4 address discovery state,
   as described in the following table:

                       a b c d e f
   1. Address literal  . . . . . .
   1. Address hostname S N . . . T
   2. auto hostname    S N . . F T
   3. auto interface   ? ? . . F ?
   3. auto socket      ? ? . . F ?
   4. auto dir header  D N D D F A

   IPv6 address discovery only uses the first IPv6 ORPort address:

                       a b c d e f
   1. ORPort listener  . . C . F .
   1. ORPort literal   . . C C F .
   1. ORPort hostname  S N C C F T

   The tables are structured as follows:
     * rows are address resolution stage variants
       * each address resolution stage has a number, and a description
       * the description includes any variants
         (for example: IP address literal, or hostname)
    * columns describe each variant's state management.

   The state management key is:
    a. What kind of API is used to perform the address resolution?
      * . a CPU-bound API
      * S a synchronous query API
      * ? an API that is probably CPU-bound, but may be synchronous on some
          platforms
      * D tor automatically updates the stored directory address, whenever a
          directory document is received
    b. What does the API depend on?
      * . a CPU-bound API
      * N a network-bound API
      * ? an API that is probably CPU-bound, but may be network-bound on some
          platforms
    c. How are any discovered addresses stored?
      * . addresses are not stored
          (but they may be cached by some higher-level tor modules)
      * D addresses are stored in the directory address suggestion variable
      * C addresses are stored in the port config listener list
    d. What event makes the address resolution happen?
      * . when tor wants to know its own address
      * D when a directory document is received
      * C when tor parses its config at startup, and during reconfiguration
    e. What conditions make tor attempt this address resolution method?
      * . this method is always attempted
      * F this method is only attempted when all other higher-priority
          methods fail to return an address
    f. Can this method timeout?
      * . can't time out
      * T might time out
      * ? probably doesn't time out, but might time out on some platforms
      * A can't time out, because it is asynchronous. If a stored address
          is available, it is returned immediately.

3.2. Finding Relay IPv6 Addresses

   We propose that relays (and bridges) try to find their IPv6 address. For
   consistency, we also propose to change the address resolution order for
   IPv4 addresses.

   We use the following general principles to choose the order of IP address
   methods:
     * Explicit is better than Implicit,
     * Local Information is better than a Remote Dependency,
     * Trusted is better than Untrusted, and
     * Reliable is better than Unreliable.
   Within these constraints, we try to find the simplest working design.

   If a relay is given the wrong address by an attacker, the attacker can
   direct all inbound relay traffic to their own address. They can't decrypt
   the traffic without the relay's private keys, but they can monitor traffic
   patterns.

   Therefore, relays should only use untrusted address discovery methods, if
   every other method has failed. Any method that uses DNS is potentially
   untrusted, because DNS is often a remote, unauthenticated service. And
   addresses provided by other directory servers are also untrusted.

   For security reasons, directory authorities only use addresses that are
   explicitly configured in their torrc.

   Based on these principles, we propose that tor tries to find relay IPv4 and
   IPv6 addresses in this order:
     1. the Address torrc option
     2. the advertised ORPort address
     3. a local interface address
        (by making an unused socket, if needed)
     4. the address of the host's own hostname (resolved using DNS, if needed)
     5. an address reported by a directory server (using X-Your-Address-Is)

   Each of these address resolution steps is described in more detail, in its
   own subsection.

   For anonymity reasons, bridges are unable to fetch directory documents over
   IPv6, until clients start to do so. (See
   [Proposal 306: Client Auto IPv6 Connections].)

   We avoid using advertised DirPorts for address resolution, because:
     * they are not supported on bridges,
     * they are not supported on IPv6,
     * they may not be configured on a relay, and
     * it is unlikely that a relay operator would configure an ORPort without
       an IPv4 address, but configure a DirPort with an IPv4 address.

   While making these changes, we want to preserve tor's existing behaviour:
     * resolve Address using the local resolver, if needed,
     * ignore private addresses on public tor networks, and
     * when there are multiple valid addresses:
       * if a list of addresses is received, choose the first address, and
       * if different addresses are received over time, choose the most recent
         address.

3.2.1. Make the Address torrc Option Support IPv6

   First, we propose that relays (and bridges) use the Address torrc option
   to find their IPv4 and IPv6 addresses.

   There are two cases we need to cover:

     1. Explicit IP addresses:
        * allow the option to be specified up to two times,
        * use the IPv4 address for IPv4,
        * use the IPv6 address for IPv6.
        Configuring two addresses in the same address family is a config error.

     2. Hostnames / DNS names:
        * allow the option to be specified up to two times,
        * look up the configured name,
        * use the first IPv4 and IPv6 address returned by the resolver, and
        Resolving multiple addresses in the same address family is not a
        runtime error, but only the first address from each family will be
        used.

   These lookups should ignore private addresses on public tor networks. If
   multiple IPv4 or IPv6 addresses are returned, the first public address from
   each family should be used.

   We should support the following combinations of address literals and
   hostnames:

   Legacy configurations:
     A. No configured Address option
     B. Address IPv4 literal
     C. Address hostname (use IPv4 and IPv6 DNS addresses)

   New configurations:
     D. Address IPv6 literal
     E. Address IPv4 literal / Address IPv6 literal
     F. Address hostname / Address hostname (use IPv4 and IPv6 DNS addresses)
     G. Address IPv4 literal / Address hostname (only use IPv6 DNS addresses)
     H. Address hostname (only use IPv4 DNS addresses) / Address IPv6 literal

   If we can't find an IPv4 or IPv6 address using the configured Address
   options:
     No IPv4: guess IPv4, and its reachability must succeed.
     No IPv6: guess IPv6, publish if reachability succeeds.

   Combinations A and B are the most common legacy configurations. We want to
   support the following outcomes for all legacy configurations:
     * automatic upgrades to guessed and reachable IPv6 addresses,
     * continuing to operate on IPv4 when the IPv6 address can't be guessed,
       and
     * continuing to operate on IPv4 when the IPv6 address has been guessed,
       but it is unreachable.

   At this time, we do not propose guessing multiple IPv4 or IPv6 addresses
   and testing their reachability (see section 3.4.2).

   It is an error to configure an Address option with a private IPv4 or IPv6
   address. Tor should warn if a configured Address hostname does not resolve
   to any publicly routable IPv4 or IPv6 addresses. (In both these cases, if
   tor is configured with a custom set of directory authorities, private
   addresses should be allowed, with a notice-level log.)

   For security reasons, directory authorities only use addresses that are
   explicitly configured in their torrc. Therefore, we propose that directory
   authorities only accept IPv4 or IPv6 address literals in their Address
   option. They must not attempt to resolve their Address using DNS. It is a
   config error to provide a hostname as a directory authority's Address.

   If the Address option is not configured for IPv4 or IPv6, or the hostname
   lookups do not provide both IPv4 and IPv6 addresses, address resolution
   should go to the next step.

3.2.2. Use the Advertised ORPort IPv4 and IPv6 Addresses

   Next, we propose that relays (and bridges) use the first advertised ORPort
   IPv4 and IPv6 addresses, as configured in their torrc.

   The ORPort address may be a hostname. If it is, tor should try to use it to
   resolve an IPv4 and IPv6 address, and open ORPorts on the first available
   IPv4 and IPv6 address. Tor should respect the IPv4Only and IPv6Only port
   flags, if specified. (Tor currently resolves IPv4 and IPv6 addresses from
   hostnames in ORPort lines.)

   Relays (and bridges) currently use the first advertised ORPort IPv6 address
   as their IPv6 address. We propose to use the first advertised IPv4 ORPort
   address in a similar way, for consistency.

   Therefore, this change may affect existing relay IPv4 addressses. We expect
   that a small number of relays may change IPv4 address, from a guessed IPv4
   address, to their first advertised IPv4 ORPort address.

   In rare cases, relays may have been using non-advertised ORPorts for their
   addresses. This change may also change their addresses.

   Tor currently uses its listener port list to look up its IPv6 ORPort for
   its descriptor. We propose that tor's address discovery uses the  listener
   port list for both IPv4 and IPv6. (And does not attempt to independently
   parse or resolve ORPort configs.)

   This design decouples ORPort option parsing, ORPort listener opening, and
   address discovery. It also implements a form of caching: IPv4 and IPv6
   addresses resolved from hostnames are stored in the listener port list,
   then used to open listeners. Therefore, tor should continue to use the same
   address, while the listener remains open. (See also sections 3.2.7 and
   3.2.8.)

   For security reasons, directory authorities only use addresses that are
   explicitly configured in their torrc. Therefore, we propose that directory
   authorities only accept IPv4 or IPv6 address literals in the address part
   of the ORPort and DirPort options. They must not attempt to resolve these
   addresses using DNS. It is a config error to provide a hostname as a
   directory authority's ORPort or DirPort.

   If directory authorities don't have an IPv4 address literal in their
   Address or ORPort, they should issue a configuration error, and refuse to
   launch. If directory authorities don't have an IPv6 address literal in their
   Address or ORPort, they should issue a notice-level log, and fall back to
   only using IPv4.

   For the purposes of address resolution, tor should ignore private
   configured ORPort addresses on public tor networks. (Binding to private
   ORPort addresses is supported, even on public tor networks, for relays that
   use NAT to reach the Internet.) If an ORPort address is private, address
   resolution should go to the next step.

3.2.3. Use Local Interface IPv6 Address

   Next, we propose that relays (and bridges) use publicly routable addresses
   from the OS interface addresses or routing table, as their IPv4 and IPv6
   addresses.

   Tor has local interface address resolution functions, which support most
   major OSes. Tor uses these functions to guess its IPv4 address. We propose
   using them to also guess tor's IPv6 address.

   We also propose modifying the address resolution order, so interface
   addresses are used before the local hostname. This decision is based
   on our principles: interface addresses are local, trusted, and reliable;
   hostname lookups may be remote, untrusted, and unreliable.

   Some developer documentation also recommends using interface addresses,
   rather than resolving the host's own hostname. For example, on recent
   versions of macOS, the man pages tell developers to use interface addresses
   (getifaddrs) rather than look up the host's own hostname (gethostname and
   getaddrinfo). Unfortunately, these man pages don't seem to be available
   online, except for short quotes (see [getaddrinfo man page] for the
   relevant quote).

   If the local interface addresses are unavailable, tor opens a UDP socket to
   a publicly routable address, but doesn't actually send any packets.
   Instead, it uses the socket APIs to discover the interface address for the
   socket. (UDP is used because it is stateless, so the OS will not send any
   packets to open a connection.)

   For security reasons, directory authorities only use addresses that are
   explicitly configured in their torrc. Since local interface addresses are
   implicit, and may depend on DHCP, directory authorities do not use this
   address resolution method (or any of the other, lower-priority address
   resolution methods).

   Relays that use NAT to reach the Internet may have no publicly routable
   local interface addresses, even on the public tor network. The NAT box has
   the publicly routable addresses, and it may be a separate machine.

   Relays may also be unable to detect any local interface addresses. The
   required APIs may be unavailable, due to:
     * missing OS or library features, or
     * local security policies.

   Tor already ignores private IPv4 interface addresses on public relays. We
   propose to also ignore private IPv6 interface addresses. If all IPv4 or
   IPv6 interface addresses are private, address resolution should go to the
   next step.

3.2.4. Use Own Hostname IPv6 Addresses

   Next, we propose that relays (and bridges) get their local hostname, look
   up its addresses, and use them as its IPv4 and IPv6 addresses.

   We propose to use the same underlying lookup functions to look up the IPv4
   and IPv6 addresses for:
     * the Address torrc option (see section 3.2.1), and
     * the local hostname.
   However, OS APIs typically only return a single hostname. (Rather than a
   separate hostname for IPv4 and IPv6.)

   For security reasons, directory authorities only use addresses that are
   explicitly configured in their torrc. Since hostname lookups may use DNS,
   directory authorities do not use this address resolution method.

   The hostname lookup should ignore private addresses on public relays. If
   multiple IPv4 or IPv6 addresses are returned, the first public address from
   each family should be used. If all IPv4 or IPv6 hostname addresses are
   private, address resolution should go to the next step.

3.2.5. Use Directory Header IPv6 Addresses

   Finally, we propose that relays get their IPv4 and IPv6 addresses from the
   X-Your-Address-Is HTTP header in tor directory documents. To support this
   change, we propose that relays start fetching directory documents over IPv4
   and IPv6.

   We propose that bridges continue to only fetch directory documents over
   IPv4, because they try to imitate clients. (Most clients only fetch
   directory documents over IPv4, a few clients are configured to only fetch
   over IPv6.) When client behaviour changes to use both IPv4 and IPv6 for
   directory fetches, bridge behaviour can also change to match. (See
   section 3.4.1 and [Proposal 306: Client Auto IPv6 Connections].)

   For security reasons, directory authorities only use addresses that are
   explicitly configured in their torrc. Since directory headers are provided
   by other directory servers, directory authorities do not use this address
   resolution method.

   We propose to use a simple load balancing scheme for IPv4 and IPv6
   directory requests:
     * choose between IPv4 and IPv6 directory requests at random.

   We do not expect this change to have any load-balancing impact on the public
   tor network, because the number of relays is much smaller than the number
   of clients. However, the 6 directory authorities with IPv6 enabled may see
   slightly more directory load, particularly over IPv6.

   To support this change, tor should also change how it handles IPv6
   directory failures on relays:
     * avoid recording IPv6 directory failures as remote relay failures,
       because they may actually be due to a lack of IPv6 connectivity on the
       local relay, and
     * issue IPv6 directory failure logs at notice level, and rate-limit them
       to one per hour.

   If a relay is:
     * explicitly configured with an IPv6 address, or
     * a publicly routable, reachable IPv6 address is discovered in an
       earlier step,
   tor should start issuing IPv6 directory failure logs at warning level. Tor
   may also record these directory failures as remote relay failures. (Rather
   than ignoring them, as described in the previous paragraph.)

   (Alternately, tor could stop doing IPv6 directory requests entirely. But we
   prefer designs where all relays behave in a similar way, regardless of their
   internal state.)

   For some more complex directory load-balancing schemes, see section 3.5.4.

   Tor already ignores private IPv4 addresses in directory headers. We propose
   to also ignore private IPv6 addresses in directory headers. If all IPv4 and
   IPv6 addresses in directory headers are private, address resolution should
   return a temporary error.

   Whenever address resolution fails, tor should warn the operator to set the
   Address torrc option for IPv4 and IPv6. (If IPv4 is available, and only
   IPv6 is missing, the log should be at notice level.) These logs may need to
   be rate-limited.

   The next time tor receives a directory header containing a public IPv4 or
   IPv6 address, tor should use that address for reachability checks. If the
   reachability checks succeed, tor should use that address in its descriptor.

   Doing relay directory fetches over IPv6 will create extra IPv6 connections
   and IPv6 bandwidth on the tor network. (See
   [Proposal 313: Relay IPv6 Statistics].) In addition, some client circuits
   may use the IPv6 connections created by relay directory fetches.

3.2.6. Disabling IPv6 Address Resolution

   Relays (and bridges) that have a reachable IPv6 address, but that address
   is unsuitable for the relay, need to be able to disable IPv6 address
   resolution.

   Based on [Proposal 311: Relay IPv6 Reachability], and this proposal, those
   relays would:
     * discover their IPv6 address,
     * open an IPv6 ORPort,
     * find it reachable,
     * publish a descriptor containing that IPv6 ORPort,
     * have the directory authorities find it reachable,
     * have it published in the consensus, and
     * have it used by clients,
   regardless of how the operator configures their tor instance.

   Currently, relays are required to have an IPv4 address. So if the guessed
   IPv4 address is unsuitable, operators can set the Address option to a
   suitable IPv4 address. But IPv6 addresses are optional, so relay operators
   may need to disable IPv6 entirely.

   We propose a new torrc-only option, AddressDisableIPv6. This option is set
   to 0 by default. If the option is set to 1, tor disables IPv6 address
   resolution, IPv6 ORPorts, IPv6 reachability checks, and publishing an IPv6
   ORPort in its descriptor.

3.2.6.1. Disabling IPv6 Address Resolution: Alternative Design

   As an alternative design, tor could change its interpretation of the
   IPv4Only flag, so that the following configuration lines disable IPv6:
   (In the absence of any non-IPv4Only  ORPort lines.)
     * ORPort 9999 IPv4Only
     * ORPort 1.1.1.1:9999 IPv4Only

   However, we believe that this is a confusing design, because we want to
   enable IPv6 address resolution on this similar, very common configuration:
     * ORPort 1.1.1.1:9999

   Therefore, we avoid this design, becuase it changes the meaning of existing
   flags and options.

3.2.7. Automatically Enabling an IPv6 ORPort

   We propose that relays (and bridges) that discover their IPv6 address,
   should open an IPv6 ORPort, and test its reachability (see
   [Proposal 311: Relay IPv6 Reachability], particularly section 4.3.1).

   The ORPort should be opened on the port configured in the relay's ORPort
   torrc option. Relay operators can use the IPv4Only and IPv6Only options
   to configure different ports for IPv4 and IPv6.

   If the ORPort is auto-detected, there will not be any specific bind
   address. (And the detected address may actually be on a NAT box, rather
   than the local machine.) Therefore, relays should attempt to bind to all
   IPv4 and IPv6 addresses (or all interfaces).

   Some operating systems expect applications to bind to IPv4 and IPv6
   addresses using separate API calls. Others don't support binding only to
   IPv4 or IPv6, and will bind to all addresses whenever there is no specified
   IP address (in a single API call). Tor should support both styles of
   networking API.

   In particular, if binding to all IPv6 addresses fails, relays should still
   try to discover their public IPv6 address, and check the reachability of
   that address. Some OSes may not support the IPV6_V6ONLY flag, but they may
   instead bind to all addresses at runtime. (The tor install may also have
   compile-time / runtime flag mismatches.)

   If both reachability checks succeed, relays should publish their IPv4 and
   IPv6 ORPorts in their descriptor.

   If only the IPv4 ORPort check succeeds, and the IPv6 address was guessed
   (rather than being explicitly configured), then relays should:
     * publish their IPv4 ORPort in their descriptor,
     * stop publishing their IPv6 ORPort in their descriptor, and
     * log a notice about the failed IPv6 ORPort reachability check.

3.2.8. Proposed Relay IPv4 and IPv6 Address State Management

   We propose that relays (and bridges) manage their IPv4 and IPv6 address
   discovery state, as described in the following table:

                       a b c d e f
   1. Address literal  . . . . . .
   1. Address hostname S N . . . T
   2. ORPort listener  . . C . F .
   2. ORPort literal   . . C C F .
   2. ORPort hostname  S N C C F T
   3. auto interface   ? ? . . F ?
   3. auto socket      ? ? . . F ?
   4. auto hostname    S N . . F T
   5. auto dir header  D N D D F A

   See section 3.1.1 for a description and key for this table. See the rest of
   section 3.2 for a detailed description of each method and variant.

   For security reasons, directory authorities only use addresses that are
   explicitly configured in their torrc. Therefore, they stop after step 2.
   (And don't use the "hostname" variants in steps 1 and 2.)

   For anonymity reasons, bridges are unable to fetch directory documents over
   IPv6, until clients start to do so. (See
   [Proposal 306: Client Auto IPv6 Connections].)

3.3. Consequential Tor Client Changes

   We do not propose any required client address resolution changes at this
   time.

   However, clients will use the updated address resolution functions to detect
   when they are on a new connection, and therefore need to rotate their TLS
   keys.

   This minor client change allows us to avoid keeping an outdated version of
   the address resolution functions, which is only for client use.

   Clients should skip address resolution steps that don't apply to them, such
   as:
     * the ORPort option, and
     * the Address option, if it becomes a relay module option.

3.4. Alternative Address Resolution Designs

   We briefly mention some potential address resolution designs, and the
   reasons that they were not used in this proposal.

   (Some designs may be proposed for future Tor versions, but are not necessary
   at this time.)

3.4.1. Future Bridge IPv6 Address Resolution Behaviour

   When clients automatically fetch directory documents via relay IPv4 and
   IPv6 ORPorts by default, bridges should also adopt this dual-stack
   behaviour. (For example, see [Proposal 306: Client Auto IPv6 Connections].)

   When bridges fetch directory documents via IPv6, they will be able to find
   their IPv6 address using directory headers (see 3.2.5).

3.4.2. Guessing Muliple IPv4 or IPv6 Addresses

   We avoid designs which guess (or configure) multiple IPv4 or IPv6
   addresses, test them all for reachability, and choose one that works.

   Using multiple addresses is rare, and the code to handle it is complex. It
   also requires careful design to avoid:
     * conflicts between multiple relays (or bridges) on the same address
       (tor allows up to 2 relays per IPv4 address),
     * relay flapping,
     * race conditions, and
     * relay address switching.

3.4.3. Rejected Address Resolution Designs

   We reject designs that try all the different address resolution methods,
   score addresses, and then choose the address with the highest score.

   These designs are a generalisation of designs that try different methods in
   a set order (like this proposal). They are more complex than required.
   Complex designs can confuse operators, particularly when they fail.

   Operators should not need complex address resolution in tor: most relay
   (and bridge) addresses are fixed, or change occasionally. And most relays
   can reliably discover their address using directory headers, if all other
   methods fail. (Bridges won't discover their IPv6 address from directory
   headers, see section 3.2.5.)

   If complex address resolution is required, it can be configured using a
   dynamic DNS name in the Address torrc option, or via the control port.

   We also avoid designs that use any addresses other than the first
   (or latest) valid IPv4 and IPv6 address. These designs are more complex, and
   they don't have clear benefits:
     * sort addresses numerically (avoid address flipping)
     * sort addresses by length, then numerically
       (also minimise consensus size)
     * store a list of previous addresses in the state file, and use the most
       recently used address that's currently available.

   Operators who want to avoid address flipping should set the Address option
   in the torrc. Operators who want to minimise the size of the consensus
   should use all-zero IPv6 host identifiers.

3.5. Optional Efficiency and Reliability Changes

   We propose some optional changes for efficiency and reliability, and
   describe their impact.

   Some of these changes may be more appropriate in future releases, or
   along with other proposed features.

   Some of these changes make tor ignore some potential IP addresses.

   Ignoring addresses risks relays having no available ORPort addresses, and
   refusing to publish their descriptor. So before we ignore any addresses, we
   should make sure that:
     * tor's other address detection methods are robust and reliable, and
     * we would prefer relays to shut down, rather than use the ignored
       address.

   As a less severe alternative, low-quality methods can be put last in the
   address resolution order. (See section 3.2.)

   If relays prefer addresses from particular sources (for example: ORPorts),
   they should try these sources regularly, so that their addresses do not
   become too old.

   If relays ignore addresses from some sources (for example: DirPorts), they
   must regularly try other sources (for example: ORPorts).

3.5.1. Using Authenticated IPv4 and IPv6 Addresses

   We propose this optional change, to improve relay (and bridge) address
   accuracy and reliability.

   Relays should try to use authenticated connections to discover their own
   IPv4 and IPv6 addresses.

   Tor supports two kinds of authenticated address information:
     * authenticated directory connections, and
     * authenticated NETINFO cells.
   See the following sections for more details.

   See also sections 3.5.2 to 3.5.4.

3.5.1.1. Authenticated Directory Connections

   We propose this optional change, to improve relay address accuracy and
   reliability. (Bridges are not affected, because they already use
   authenticated directory connections, just like clients.)

   Tor supports authenticated, encrypted directory fetches using BEGINDIR over
   ORPorts (see the [Tor Specification] for details).

   Relays currently fetch unencrypted directory documents over DirPorts. The
   directory document itself is signed, but the HTTP headers are not
   authenticated. (Clients and bridges only fetch directory documents using
   authenticated directory fetches.)

   Using authenticated directory headers for relay addresses:
     * provides authenticated address information,
     * reduces the number of attackers that can deliberately give a relay an
       incorrect IP address, and
     * avoids caches (or other machines) accidentally mangling, deleting, or
       repeating X-Your-Address-Is headers.

   To make this change, we need to modify tor's directory connection code:
     * when making directory requests, relays should fetch some directory
       documents using BEGINDIR over ORPorts.

   Once tor regularly gets authenticated X-Your-Address-Is headers, relays can
   change how they handle unauthenticated addresses. When they receive an
   unauthenticated address suggestion, relays can:
     * ignore the address, or
     * use the address as the lowest priority address method.
   See section 3.5 for some factors to consider when making this design
   decision.

   For security reasons, directory authorities only use addresses that are
   explicitly configured in their torrc. Since directory headers are provided
   by other directory servers, directory authorities do not use this address
   resolution method.

   For anonymity reasons, bridges are unable to fetch directory documents over
   IPv6, until clients start to do so. (See
   [Proposal 306: Client Auto IPv6 Connections].)

   Bridges currently use authenticated IPv4 connections for all their
   directory fetches, to imitate default client behaviour.

   We describe a related change, which is also optional:

   We can increase the number of ORPort directory fetches:
     * if tor has an existing ORPort connection to a relay that it has selected
       for a directory fetch, it should use an ORPort fetch, rather than
       opening an additional DirPort connection.

   Using an existing ORPort connection:
     * saves one DirPort connection and file descriptor,
     * but slightly increases the cryptographic processing done by the relay,
       and by the directory server it is connecting to.
   However, the most expensive cryptographic operations have already happened,
   when the ORPort connection was opened.

   This change does not increase the number of NETINFO cells, because it
   re-uses existing OR connections. See the next section for more details.

3.5.1.2. Authenticated NETINFO Cells

   We propose this optional change, to improve relay (and bridge) address
   accuracy and reliability. (Bridge IPv6 addresses are not affected, because
   bridges only make OR connections over IPv4, to imitate default client
   behaviour.)

   Tor supports authenticated IPv4 and IPv6 address information, using the
   NETINFO cells exchanged at the beginning of each ORPort connection (see the
   [Tor Specification] for details).

   Relays do not currently use any address information from NETINFO cells.

   Using authenticated NETINFO cells for relay addresses:
     * provides authenticated address information,
     * reduces the number of attackers that can deliberately give a relay an
       incorrect IP address, and
     * does not require a directory fetch (NETINFO cells are sent during
       connection setup).

   To make this change, we need to modify tor's cell processing:
     * when processing NETINFO cells, tor should store the OTHERADDR field,
       like it currently does for X-Your-Address-Is HTTP headers, and
     * IPv4 and IPv6 addresses should be stored separately.
   See the previous section, and section 3.2.5 for more details about the
   X-Your-Address-Is HTTP header.

   Once tor uses NETINFO cell addresses, relays can change how they handle
   unauthenticated X-Your-Address-Is headers. When they receive an
   unauthenticated address suggestion, relays can:
     * ignore the address, or
     * use the address as the lowest priority address method.
   See section 3.5 for some factors to consider when making this design
   decision.

   We propose that tor continues to use the X-Your-Address-Is header, and adds
   support for addresses in NETINFO cells. X-Your-Address-Is headers are sent
   once per directory document fetch, but NETINFO cells are only sent once per
   OR connection.

   If a relay:
     * only gets addresses from NETINFO cells from authorities, and
     * has an existing, long-term connection to every authority,
   then it may struggle to detect address changes.

   Once all supported tor versions use NETINFO cells for address detection, we
   should review this design decision. If we are confident that almost all
   relays will be forced to make new connections when their address changes,
   then tor may be able to stop using X-Your-Address-Is HTTP headers.

   For security reasons, directory authorities only use addresses that are
   explicitly configured in their torrc. Since NETINFO cells are provided
   by other directory servers, directory authorities do not use this address
   resolution method.

   Bridges only make OR connections, and those OR connections are only over
   IPv4, to imitate default client behaviour.

   For anonymity reasons, bridges are unable to make regular connections over
   IPv4 and IPv6, until clients start to do so. (See
   [Proposal 306: Client Auto IPv6 Connections].)

   As an alternative design, if tor's addresses are stale, it could close some
   of its open directory authority connections. (Similar to section 4.4.2
   in [Proposal 311: Relay IPv6 Reachability], where relays close existing OR
   connections, before testing their own reachability.) However, this design is
   more complicated, because it involves tracking address age, as well as the
   address itself.

3.5.2. Preferring IPv4 and IPv6 Addresses from Directory Authorities

   We propose this optional change, to improve relay (but not bridge) address
   accuracy and reliability.

   Relays prefer IPv4 and IPv6 address suggestions received from Directory
   Authorities.

   Directory authorities do not use these address detection methods to
   discover their own addresses, for security reasons.

   When they receive an address suggestion from a directory mirror, relays can:
     * ignore the address, or
     * use the address as the lowest priority address method.
   See section 3.5 for some factors to consider when making this design
   decision.

   Bridges only make OR connections, and those OR connections are only over
   IPv4, to imitate default client behaviour.

   For anonymity reasons, bridges are unable to make regular connections over
   IPv6, until clients start to do so. (See
   [Proposal 306: Client Auto IPv6 Connections].)

   See also sections 3.5.1 to 3.5.4.

3.5.3. Ignoring Addresses on Inbound Connections

   We propose this optional change, to improve relay (and bridge) address
   accuracy and reliability.

   Relays ignore IPv4 and IPv6 address suggestions received on inbound
   connections.

   We make this change, because we want to detect the IP addresses of the
   relay's outbound routes, rather than the addresses that that other relays
   believe they are connecting to for inbound connections.

   If we make this change, relays may need to close some inbound connections,
   before doing address detection. If we also make the changes in sections
   3.5.1 and 3.5.2, busy relays could have persistent, inbound OR connections
   from all directory authorities. (Currently, there are 9 directory
   authorities with IPv4 addresses, and 6 directory authorities with IPv6
   addresses.)

   Directory authorities do not use these address detection methods to
   discover their own addresses, for security reasons.

   See also sections 3.5.1 to 3.5.4.

3.5.4. Load Balancing

   We propose some optional changes to improve relay (and bridge)
   load-balancing across directory authorities.

   Directory authorities do not use these address detection methods to
   discover their own addresses, for security reasons.

   See also sections 3.5.1 to 3.5.3.

3.5.4.1. Directory Authority Load Balancing

   Relays may prefer:
     * authenticated connections (section 3.5.1).

   Relays and bridges may prefer:
     * connecting to Directory Authorities (section 3.5.2), or
     * ignoring addresses on inbound connections (section 3.5.3)
       (and therefore, they may close some inbound connections,
       leading to extra connection re-establishment load).

   All these changes are optional, so they might not be implemented.

   Directory authorities do not use these address detection methods to
   discover their own addresses, for security reasons.

   If both changes are implemented, we would like all relays (and bridges) to
   do frequent directory fetches:
     * using BEGINDIR over ORPorts,
     * to directory authorities.
   However, this extra load from relays may be unsustainable during high
   network load (see
   [Ticket 33018: Dir auths using an unsustainable 400+ mbit/s]).

   For anonymity reasons, bridges should avoid connecting to directory
   authorities too frequently, to imitate default client behaviour.

   Therefore, we propose a simple load-balancing scheme between address
   resolution and non-address resolution requests:
     * when relays first start up, they should make two directory authority
       ORPort fetch attempts, one on IPv4, and one on IPv6,
     * relays should also make occasional directory authority ORPort directory
       fetch attempts, on IPv4 and IPv6, to learn if their addresses have
       changed.

   We propose a new torrc option and consensus parameter:

    RelayMaxIntervalWithoutAddressDetectionRequest N seconds|minutes|hours

    Relays make most of their directory requests via directory mirror DirPorts,
    to reduce the load on directory authorities.

    When this amount of time has passed since a relay last connected to a
    directory authority ORPort, the relay makes its next directory request via
    a directory authority ORPort. (Default: 15 minutes)

   The final name and description for this option will depend on which optional
   changes are actually implemented in tor. In particular, this option should
   only consider requests that tor may use to discover its IP addresses.
   For example:
     * if tor uses NETINFO cells for addresses (section 3.5.1.2), then all
       OR connections to an authority should be considered,
     * if tor does not use NETINFO cells for addresses, and only uses
       X-Your-Address-Is headers, then only directory fetches from authorities
       should be considered.

   We set the default value of this option to 15 minutes, because:
     * tor's reachability tests fail if the ORPort is unreachable after 20
       minutes. So we want to do at least two address detection requests in
       the first 20 minutes;
     * the minimum consensus period is 30 minutes, and we want to do at least
       one address detection per consensus period. (Consensuses are usually
       created every hour. But if there is no fresh consensus, directory
       authorities will try to create a consensus every 30 minutes); and
     * the default value for TestingAuthDirTimeToLearnReachability is 30
       minutes. So directory authorities will make reachability test OR
       connections to each relay, at least every 30 minutes. Therefore, relays
       will see NETINFO cells from directory authorities about this often.
       (Relays may use NETINFO cells for address detection, see section
       3.5.1.2.)

   See also section 3.5.4.3, for some general load balancing criteria, that
   may help when tuning the address detection interval.

   We propose a related change, which is also optional:

   If relays use address suggestions from directory mirrors, they may choose
   between ORPort and DirPort connections to directory mirrors at random.
   Directory mirrors typically have enough spare CPU and bandwidth to handle
   ORPort directory requests. (And the most expensive cryptography happens
   when the ORPort connection is opened.)

   See also sections 3.5.1 to 3.5.3.

3.5.4.2. Load Balancing Between IPv4 and IPv6 Directories

   We propose this optional change, to improve the load-balancing between IPv4
   and IPv6 directories, when used by relays to find their IPv4 and IPv6
   addresses (see section 3.2.5).

   For anonymity reasons, bridges are unable to make regular connections over
   IPv6, until clients start to do so. (See
   [Proposal 306: Client Auto IPv6 Connections].)

   Directory authorities do not use these address detection methods to
   discover their own addresses, for security reasons.

   This change may only be necessary if the following changes result in poor
   load-balancing, or other relay issues:
     * randomly selecting IPv4 or IPv6 directories (see section 3.2.5),
     * preferring addresses from directory authorities, via an authenticated
       connection (see sections 3.5.1 and 3.5.2), or
     * ignoring addresses on inbound connections, and therefore closing and
       re-opening some connections (see section 3.5.3).

   We propose that the RelayMaxIntervalWithoutAddressDetection option is
   counted separately for IPv4 and IPv6 (see the previous section for details).

   For example:
     * if 30 minutes has elapsed since the last IPv4 address detection request,
       then the next directory request should be an IPv4 address detection
       request, and
     * if 30 minutes has elapsed since the last IPv6 address detection request,
       then the next directory request should be an IPv6 address detection
       request.

   If both intervals have elapsed at the same time, the relay should choose
   between IPv4 and IPv6 at random.

   See also section 3.5.4.3, for some general load balancing criteria, that
   may help when tuning the address detection interval.

   Alternately, we could wait until
   [Proposal 306: Client Auto IPv6 Connections] is implemented, and use the
   directory fetch design from that proposal.

   See also sections 3.5.1 to 3.5.3.

3.5.4.3. General Load Balancing Criteria

   We propose the following criteria for choosing load-balancing intervals:

   The selected interval should be chosen based on the following factors:
     * relays need to discover their IPv4 and IPv6 addresses to publish their
       descriptors,
     * it only takes one successful directory fetch from one authority for a
       relay to discover its IP address (see section 3.5.2),
     * if relays fall back to addresses discovered from directory mirrors,
       when directory authorities are unavailable (see section 3.5.2),
     * BEGINDIR over ORPort requires and TLS connection, and some additional
       tor cryptography, so it is more expensive for authorities than a
       DirPort fetch (and it can not be cached by a HTTP cache)
       (see section 3.5.1),
     * closing and re-opening some OR connections (see section 3.5.3),
     * minimising wasted CPU (and bandwidth) for IPv6 connection attempts on
       IPv4-only relays, and
     * other potential changes to relay directory fetches (see
       [Ticket 33018: Dir auths using an unsustainable 400+ mbit/s])

   The selected interval should allow almost all relays to update both their
   IPv4 and IPv6 addresses:
     * at least twice when they bootstrap and test reachability (to allow for
       fetch failures),
     * at least once per consensus interval (that is, every 30 minutes), and
     * from a directory authority (if required).

   For anonymity reasons, bridges are unable to make regular connections over
   IPv6, until clients start to do so. (See
   [Proposal 306: Client Auto IPv6 Connections].)

   Directory authorities do not use these address detection methods to
   discover their own addresses, for security reasons.

   In this proposal, relays choose between IPv4 and IPv6 directory fetches
   at random (see section 3.2.5 for more detail). But if this change causes
   issues on IPv4-only relays, we may have to try IPv6 less often.

   See also sections 3.5.1 to 3.5.3.

3.5.5. Detailed Address Resolution Logs

   We propose this optional change, to help diagnose relay address resolution
   issues.

   Relays (and bridges) should  log the address chosen using each address
   resolution method, when:
     * address resolution succeeds,
     * address resolution fails,
     * reachability checks fail, or
     * publishing the descriptor fails.
   These logs should be rate-limited separately for successes and failures.

   The logs should tell operators to set the Address torrc option for IPv4 and
   IPv6 (if available).

3.5.6. Add IPv6 Support to is_local_addr()

   We propose this optional change, to improve the accuracy of IPv6 address
   detection from directory documents.

   Directory servers use is_local_addr() to detect if the requesting tor
   instance is on the same local network. If it is, the directory server does
   not include the X-Your-Address-Is HTTP header in directory documents.

   Currently, is_local_addr() checks for:
     * an internal IPv4 or IPv6 address, or
     * the same IPv4 /24 as the directory server.

   We propose also checking for:
     * the same IPv6 /48 as the directory server.

   We choose /48 because it is typically the smallest network in the global
   IPv6 routing tables, and it was previously the recommended per-customer
   network block. (See [RFC 6177: IPv6 End Site Address Assignment].)

   Tor currently uses:
     * IPv4 /8 and IPv6 /16 for port summaries,
     * IPv4 /16 and IPv6 /32 for path selection (avoiding relays in the same
       network block).
   See also the next section, which uses IPv6 /64 for sybils.

3.5.7. Add IPv6 Support to AuthDirMaxServersPerAddr

   We propose this optional change, to improve the health of the network, by
   rejecting too many relays on the same IPv6 address.

   Modify get_possible_sybil_list() so it takes an address family argument,
   and returns a list of IPv4 or IPv6 sybils.

   Use the modified get_possible_sybil_list() to exclude relays from the
   authority's vote, if there are more than:
     * AuthDirMaxServersPerAddr on the same IPv4 address, or
     * AuthDirMaxServersPerIPv6Site in the same IPv6 /64.

   We choose IPv6 /64 as the IPv6 site size, because:
     * provider site allocations range between /48 and /64
       (with a recommendation of /56),
     * /64 is the typical host allocation
       (see [RFC 6177: IPv6 End Site Address Assignment]),
     * we don't want to discourage IPv6 address adoption on the tor network.

   Tor currently uses:
     * IPv4 /8 and IPv6 /16 for port summaries,
     * IPv4 /16 and IPv6 /32 for path selection (avoiding relays in the same
       network block).
   See also the previous section, which uses IPv6 /48 for the local network.

   This change allows:
     * up to AuthDirMaxServersPerIPv6Site relays on the smallest IPv6 site
       (/64, which is also the typical IPv6 host), and
     * thousands of relays on the recommended IPv6 site size of /56.
   The number of relays in an IPv6 block was previously unlimited, and sybils
   were only limited by the scarcity of IPv4 addresses.

   We propose choosing a default value for AuthDirMaxServersPerIPv6Site by
   analysing the current IPv6 addresses on the tor network. Reasonable
   default values are likely in the range 4 to 50.

   If tor every allows IPv6-only relays, we should review the default value
   of AuthDirMaxServersPerIPv6Site.

   Since these relay exclusions happen at voting time, they do not require a
   new consensus method.

3.5.8. Use a Local Interface Address on the Default Route

   We propose this optional change, to improve the accuracy of local interface
   IPv4 and IPv6 address detection (see section 3.2.3), on relays
   (and bridges).

   Directory authorities do not use this address detection method to
   discover their own addresses, for security reasons.

   Rewrite the get_interface_address*() functions to choose an interface
   address on the default route, or to sort default route addresses first in
   the list of addresses. (If the platform API allows us to find the default
   route.)

   For more information, see [Ticket 12377: Prefer default route when checking
   local interface addresses].

   This change might not be necessary, because the directory header IP address
   method will find the IP address of the default route, in most cases
   (see section 3.2.5).

3.5.9. Add IPv6 Support via Other DNS APIs

   We propose these optional changes, to add IPv6 support to hostname
   resolution on older OSes. These changes affect:
     * the Address torrc option, when it is a hostname (see section 3.2.1),
       and
     * automatic hostname resolution (see section 3.2.4),
   on relays and bridges.

   Directory authorities do not use this address detection method to
   discover their own addresses, for security reasons.

   Tor currently uses getaddrinfo() on most systems, which supports IPv6 DNS.
   But tor also supports the legacy gethostbyname() DNS API, which does not
   support IPv6.

   There are two alternative APIs we could use for IPv6 DNS, if getaddrinfo()
   is not available:
     * libevent DNS API, and
     * gethostbyname2().

   But this change may be unnecessary, because:
     * Linux has used getaddrinfo() by default since glibc 2.20 (2014)
     * macOS has recommended getaddrinfo() since before 2006
     * since macOS adopts BSD changes, most BSDs would have switched to
       getaddrinfo() in a similar timeframe
     * Windows has supported getaddrinfo() since Windows Vista; tor's minimum
       supported Windows version is Vista.
   See [Tor Supported Platforms] for more detai

   If a large number of systems do not support getaddrinfo(), we propose
   implementing one of these alternatives:

   The libevent DNS API supports IPv6 DNS, and tor already has a dependency on
   libevent. Therefore, we should prefer the libevent DNS API. (Unless we find
   it difficult to implement.)

   We could also use gethostbyname2() to add IPv6 support to hostname
   resolution on older OSes, which don't support getaddrinfo().

   Handling multiple addresses:

   When looking up hostnames using libevent, the DNS callbacks provide a list
   of all addresses received. Therefore, we should ignore any private
   addresses, and then choose the first address in the list.

   When looking up hostnames using gethostbyname() or gethostbyname2(), if the
   first address is a private address, we may want to look at the entire list
   of addresses. Some struct hostent versions (example: current macOS) also
   have a h_addr_list rather than h_addr. (They define h_addr as
   h_addr_list[0], for backwards compatibility.)

   However, having private and public addresses resolving from the same
   hostname is a rare configuration, so we might not need to make this change.
   (On OSes that support getaddrinfo(), tor searches the list of addresses for
   a publicly routable address.)

   Alternative change: remove gethostbyname():

   As an alternative, if we believe that all supported OSes have getaddrinfo(),
   we could simply remove the gethostbyname() code, rather than trying to
   modify it to work with IPv6.

   Most relays can reliably discover their address using directory headers,
   if all other methods fail. Or operators can set the Address torrc option to
   an IPv4 or IPv6 literal.

3.5.10. Change Relay OutboundBindAddress Defaults

   We propose this optional change, to improve the reliability of
   IP address-based filters in tor. These filters typically affect relays and
   directory authorities. But we propose that bridges and clients also make
   this change, for consistency.

   For example, the tor network treats relay IP addresses differently when:
     * resisting denial of service, and
     * selecting canonical, long-term connections.
   (See [Ticket 33018: Dir auths using an unsustainable 400+ mbit/s] for the
   initial motivation for this change: resisting significant bandwidth load
   on directory authorities.)

   Now that tor knows its own addresses, we propose that relays (and bridges)
   set their IPv4 and IPv6 OutboundBindAddress to these discovered addresses,
   by default. If binding fails, tor should fall back to an unbound socket.

   Operators would still be able to set a custom IPv4 and IPv6
   OutboundBindAddress, if needed.

   Currently, tor doesn't bind to a specific address, unless
   OutboundBindAddress is configured. So on relays with multiple IP addresses,
   the outbound address comes from the chosen route for each TCP connection
   or UDP packet (usually the default route).

3.5.11. IPv6 Address Privacy Extensions

   We propose this optional change, to improve the reliability of relays (and
   bridges) that use IPv6 address privacy extensions (see section 3.5 of
   [RFC 4941: Privacy Extensions for IPv6]).

   Directory authorities:
     * should not use IPv6 address privacy extensions, because their addresses
       need to stay the same over time, and
     * do not use address detection methods that would automatically select
       an IPv6 address with privacy extensions, for security reasons.

   We propose that tor should avoid using IPv6 addresses generated using
   privacy extensions, unless no other publicly routable addresses are
   available.

   In practice, each operating system has a different way of detecting IPv6
   address privacy extensions. And some operating systems may not tell
   applications if a particular address is using privacy extensions. So
   implementing this change may be difficult.

   On operating systems that provide IPv6 address privacy extension state,
   IPv6 addresses may be:
     * "public" - these addresses do not change
     * "temporary" - these addresses change due to IPv6 privacy extensions.
   Therefore, tor should prefer "public" IPv6 addresses, when they are
   available.

   However, even if we do not make this change, tor should be compatible with
   the RFC 4941 defaults:
     * a new IPv6 address is generated each day
     * deprecated addresses are removed after one week
     * temporary addresses should be disabled, unless an application opts in
       to using them
   (See sections 3.5 and  3.6 of [RFC 4941: Privacy Extensions for IPv6].)

   In particular, it can take up to 4.5 hours for a client to receive a new
   address for a relay. Here are the maximum times:
     * 30 minutes for directory authorities to do reachability checks
       (see TestingAuthDirTimeToLearnReachability in the [Tor Manual Page]).
     * 1 hour for a reachable relay to be included in a vote
     * 10 minutes for votes to be turned into a consensus
     * 2 hours and 50 minutes for clients
       (See the [Tor Directory Protocol], sections 1.4 and 5.1, and the
       corresponding Directory Authority options in the [Tor Manual Page].)

   But 4.5 hours is much less than 1 week, and even significantly less than 1
   day. So clients and relays should be compatible with the IPv6 privacy
   extensions defaults, even if they are used for all applications.

   However, bandwidth authorities may reset a relay's bandwidth when its IPv6
   address changes. (The tor network currently uses torflow and sbws as
   bandwidth authorities, neither implementation resets bandwidth when IPv6
   addresses change.) Since bandwidth authorities only scan the whole tor
   network about once a day, resetting a relay's bandwidth causes a huge
   penalty.

   Therefore, we propose that sbws should not reset relay bandwidths when
   IPv6 addresses change. (See
   [Ticket 28725: Reset relay bandwidths when their IPv6 address changes].)

3.5.12. Quick Extends After Relay Restarts

   We propose this optional change, to reduce client circuit failures, after a
   relay restarts.

   We propose that relays (and bridges) should open their ORPorts, and support
   client extends, as soon as possible after they start up. (Clients may
   already have the relay's addresses from a previous consensus.)

   Automatically enabling an IPv6 ORPort creates a race condition with IPv6
   extends (see section 3.2.7 of this proposal, and
   [Proposal 311: Relay IPv6 Reachability]).

   This race condition has the most impact when:
     1. a relay has outbound IPv6 connectivity,
     2. the relay detects a publicly routable IPv6 address,
     3. the relay opens an IPv6 ORPort,
     4. but the IPv6 ORPort is not reachable.

   Between steps 3 and 4, the relay could successfully extend over IPv6, even
   though its IPv6 ORPort is unreachable. However, we expect this case to be
   rare.

   A far more common case is that a working relay has just restarted, and
   clients still have its addresses, therefore they continue to try to extend
   through it. If the relay refused to extend, all these clients would have to
   retry their circuits.

   To support this case, tor relays should open IPv4 and IPv6 ORPorts, and
   perform extends, as soon as they can after startup. Relays can extend to
   other relays, as soon as they have validated the directory documents
   containing other relays' public keys.

   In particular, relays which automatically detect their IPv6 address, should
   support IPv6 extends as soon as they detect an IPv6 address. (Relays may
   also attempt to bind to all IPv6 addresses on all interfaces. If that bind
   is successful, they may choose to extend over IPv6, even before they know
   their own IPv6 address.)

   Relays should not wait for reachable IPv4 or IPv6 ORPorts before they start
   performing client extends.

   DirPort requests are less critical, because relays and clients will retry
   directory fetches using multiple mirrors. However, DirPorts may also open
   as early as possible, for consistency. (And for simpler code.)

   Tor's existing code handles this use case, so the code changes required to
   support IPv6 may be quite small. But we should still test this use case for
   clients connecting over IPv4 and IPv6, and extending over IPv4 and IPv6.

   Directory authorities do not rely on their own reachability checks, so they
   should be able to perform extends (and serve cached directory documents)
   shortly after startup.

3.5.13. Using Authority Addresses for Socket-Based Address Detection

   We propose this optional change, to avoid issues with firewalls during
   relay (and bridge) address detection. (And to reduce user confusion about
   firewall notifications which show a strange IP address, particularly on
   clients.)

   Directory authorities do not use a UDP socket to discover their own
   addresses, for security reasons. Therefore, we are free to use any
   directory address for this check, without the risk of a directory authority
   making a UDP socket to itself, and discovering its own private address.

   We propose that tor should use a directory authority IPv4 and IPv6 address,
   for any sockets that it opens to detect local interface addresses (see
   section 3.2.3). We propose that this change is applied regardless of the
   role of the current tor instance (relay, bridge, directory authority, or
   client).

   Tor currently uses the arbitrary IP addresses 18.0.0.1 and [2002::], which
   may be blocked by firewalls. These addresses may also cause user confusion,
   when they appear in logs or notifications.

   The relevant function is get_interface_address6_via_udp_socket_hack() in
   lib/net. The hard-coded addresses are in app/config. Directly using these
   addresses would break tor's module layering rules, so we propose:
     * copying one directory authority's hard-coded IPv4 and IPv6 addresses to
       an ADDRESS_PRIVATE macro or variable in lib/net/address.h
     * writing a unit test that makes sure that the address used by
       get_interface_address6_via_udp_socket_hack() is still in the list of
       hard-coded directory authority addresses.

   When we choose the directory authority, we should avoid using a directory
   authority that has different hard-coded and advertised IP addresses. (To
   avoid user confusion.)

4. Directory Protocol Specification Changes

   We propose explicitly supporting IPv6 X-Your-Address-Is HTTP headers in the
   tor directory protocol.

   We propose the following changes to the [Tor Directory Protocol]
   specification, in section 6.1:

  Servers MAY include an X-Your-Address-Is: header, whose value is the
  apparent IPv4 or IPv6 address of the client connecting to them. IPv6
  addresses SHOULD/MAY (TODO) be formatted enclosed in square brackets.

  TODO: require brackets? What does Tor currently do?

  For directory connections tunneled over a BEGIN_DIR stream, servers SHOULD
  report the IP from which the circuit carrying the BEGIN_DIR stream reached
  them.

  Servers SHOULD disable caching of multiple network statuses or multiple
  server descriptors.  Servers MAY enable caching of single descriptors,
  single network statuses, the list of all server descriptors, a v1
  directory, or a v1 running routers document, with appropriate expiry times
  (around 30 minutes). Servers SHOULD disable caching of X-Your-Address-Is
  headers.

5. Test Plan

   We provide a quick summary of our testing plans.

5.1. Testing Relay IPv6 Addresses Discovery

   We propose to test these changes using chutney networks. However, chutney
   creates a limited number of configurations, so we also need to test these
   changes with relay operators on the public network.

   Therefore, we propose to test these changes on the public network with a
   small number of relays and bridges.

   Once these changes are merged, volunteer relay and bridge operators will be
   able to test them by:
     * compiling from source,
     * running nightly builds, or
     * running alpha releases.

5.2. Test Existing Features

   We will modify and test these existing features:
     * Find Relay IPv4 Addresses

   We do not plan on modifying these existing features:
     * relay address retries
     * existing warning logs
   But we will test that they continue to function correctly, and fix any bugs
   triggered by the modifications in this proposal.

6. Ongoing Monitoring

   To monitor the impact of these changes:
     * relays should collect basic IPv6 connection statistics, and
     * relays and bridges should collect basic IPv6 bandwidth statistics.
   (See [Proposal 313: Relay IPv6 Statistics]).

   Some of these statistics may be included in tor's heartbeat logs, making
   them accessible to relay operators.

   We do not propose to collect additional statistics on:
     * circuit counts, or
     * failure rates.
   Collecting statistics like these could impact user privacy.

   We also plan to write a script to calculate the number of IPv6 relays in
   the consensus. This script will help us monitor the network during the
   deployment of these new IPv6 features.

7. Changes to Other Proposals

   [Proposal 306: Client Auto IPv6 Connections] needs to be modified to keep
   bridge IPv6 behaviour in sync with client IPv6 behaviour. (See section
   3.2.5.)

References:

[getaddrinfo man page]: See the quoted section in:
   https://stackoverflow.com/a/42351676

[Proposal 306: Client Auto IPv6 Connections]: One possible design for
   automatic client IPv4 and IPv6 connections is at
   https://gitweb.torproject.org/torspec.git/tree/proposals/306-ipv6-happy-eyeballs.txt
   (TODO: modify to include bridge changes with client changes)

[Proposal 311: Relay IPv6 Reachability]:
   https://gitweb.torproject.org/torspec.git/tree/proposals/311-relay-ipv6-reachability.txt

[Proposal 313: Relay IPv6 Statistics]:
   https://gitweb.torproject.org/torspec.git/tree/proposals/313-relay-ipv6-stats.txt

[RFC 4941: Privacy Extensions for IPv6]:
   https://tools.ietf.org/html/rfc4941
   Or the older RFC 3041: https://tools.ietf.org/html/rfc3041

[RFC 6177: IPv6 End Site Address Assignment]:
   https://tools.ietf.org/html/rfc6177#page-7

[Ticket 12377: Prefer default route when checking local interface addresses]:
   https://trac.torproject.org/projects/tor/ticket/12377

[Ticket 28725: Reset relay bandwidths when their IPv6 address changes]:
   https://trac.torproject.org/projects/tor/ticket/29725#comment:3

[Ticket 33018: Dir auths using an unsustainable 400+ mbit/s]:
   https://trac.torproject.org/projects/tor/ticket/33018

[Tor Directory Protocol]:
   (version 3) https://gitweb.torproject.org/torspec.git/tree/dir-spec.txt

[Tor Manual Page]:
   https://2019.www.torproject.org/docs/tor-manual.html.en

[Tor Specification]:
   https://gitweb.torproject.org/torspec.git/tree/tor-spec.txt

[Tor Supported Platforms]:
   https://trac.torproject.org/projects/tor/wiki/org/teams/NetworkTeam/SupportedPlatforms#OSSupportlevels