[tor-bugs] #17217 [Tor]: Change clients to automatically use IPv6 if they can bootstrap over it

[Tor Bug Tracker & Wiki]

#17217: Change clients to automatically use IPv6 if they can bootstrap over it
-------------------------------------+------------------------------------
 Reporter:  teor                     |          Owner:
     Type:  enhancement              |         Status:  new
 Priority:  Low                      |      Milestone:  Tor: 0.2.8.x-final
Component:  Tor                      |        Version:
 Severity:  Major                    |     Resolution:
 Keywords:  TorCoreTeam201601, ipv6  |  Actual Points:
Parent ID:  #17811                   |         Points:  medium
  Sponsor:                           |
-------------------------------------+------------------------------------
Changes (by teor):

 * keywords:  TorCoreTeam201512, ipv6 => TorCoreTeam201601, ipv6
 * priority:  Medium => Low
 * severity:  Normal => Major


Comment:

 See #17835 for details of how we could automatically determine a preferred
 IP version.

 This ticket could implement a IPv6 autodetection scheme based on #17835:
 * Set ClientUseIPv4 to 1 (for backwards-compatibility) and ClientUseIPv6
 to auto by default.
 * If the client has any non-localhost, non-link-local IPv6 addresses, act
 as if ClientUseIPv6 is 1
 * Check regularly if the interface addresses have changed. (Relays check
 every 20 minutes, clients, especially mobile/wireless clients, might want
 to check more frequently.)

 Open Questions:
  * Do we want to ignore private IPv6 addresses if the client has a public
 IPv4 address? (Do people route via private IPv6 addresses to public
 networks? Do people do IPv6 NAT?)
   * If we do, we risk clients which are on strange setups (private IPv6
 that works, public IPv4 that doesn't) failing to bootstrap.
   * If we don't, we might slow down IPv4 clients with public IPv4
 addresses, with private IPv6 addresses that don't go anywhere.
 * Is it OK to slow down all (IPv4) clients for the sake of those that can
 connect only/better via IPv6?
   * Tor Browser already sets ClientUseIPv6 to 1, so the answer is probably
 "yes".
   * We should be able to tune the (multiple concurrent) connection
 schedules in #4483 to help with this, at least during bootstrap.

 I'm marking this as low priority, to be resolved in January, because Tor
 Browser already sets ClientUseIPv6 to 1 by default. One simple resolution
 to this issue is to follow Tor Browser's lead, and set ClientUseIPv6 to
 default to 1 in tor.

 (If we're going to make IPv6 bootstrap work in 0.2.8, we also need to
 resolve this issue in the same timeframe.)

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/17217#comment:3>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online