Date: Mon, 01 Jun 2015 19:20:32 +0000 From: nusenu <nusenu@openmailbox.org>

Hi,

by comparing different methodologies of "parsing" myfamily data I stumbled upon differences between onionoo and compass.

After manual review I assume there is a bug in onionoo (or onionoo has a different opinion on what families actually are)

Example:

According to onionoo, torpidsDEevanzo [1] is part of a family with 38 members.

It lists torpidsFRonline [2] as one of its members, that implies that torpidsFRonline lists torpidsDEevanzo as one of its members as well, but torpidsDEevanzo does _not_ list torpidsFRonline (according to onionoo data).

grep 'fingerprint":"0C77421C890D16B6D201283A2' details.json|grep 5510FC1736B16D46D3F2DDA5011995C478D42594 (no result)

Is this a bug?

thanks

[1] https://atlas.torproject.org/#details/5510FC1736B16D46D3F2DDA5011995C478D425...

[2] https://atlas.torproject.org/#details/0C77421C890D16B6D201283A2244F43DF5BC89...

No, it's a feature :-) MyFamily requires bidirectional declarations to be effective. This prevents a malicious relay nominating significant portions of the Tor network as its family, in order to direct traffic to another malicious relay. (And/or slowing down the network and attempting to cause a DoS.) In this case: torpids relays have inconsistent MyFamily configurations. OnionOO appears to correctly implement the bidirectional MyFamily logic, and remove inconsistent one-way MyFamily declarations. Compass appears to believe each relay's MyFamily claims, without checking the other relay. This appears to be a fairly harmless bug in Compass, as Compass itself is not used for path selection. teor teor2345 at gmail dot com pgp 0xABFED1AC https://gist.github.com/teor2345/d033b8ce0a99adbc89c5 teor at blah dot im OTR D5BE4EC2 255D7585 F3874930 DB130265 7C9EBBC7