Test network

Posted on Oct 6, 2025

Test Network Session

Several members of the community have expressed interest in the existence or construction of a test network for tor. Such a network existed in the past but has since vanished.

Situation of Test Network at TPI

TPI has need for a test network for testing transition to arti relay. For this purpose, there is a ticket (P141) and funding on their side to 2027. The plan is to run a ctor network and slowly transition to arti relay and find out what fails. We don’t know the latest updates as the status update meeting was during TCG buildup.

Community involvement is planned for the following reasons:

  • for the community to gain experience with arti relay
  • for community to adjust deployment methods (ansible, …)

TPI requirements are likely:

  • network participants need to be reachable within a reasonable time-limit
  • network participants may need to run patched software
  • metrics.tp, collector.tp

Community Test Network

We think a community test network, optionally in parallel to an arti-specific test network would be helpful for the tor community for the following use cases:

  • arti testing
  • ipv6 testing
  • transparent consensus testing
  • allow new operators to safely test their setups
  • “lan only” network for local experiments
  • allow others to run directory authorities without the trust requirements
  • allow development of metrics and alerts
  • foster community through more direct collaboration
  • general research

Community Requirements / Ideas

  • How public should the test network be?
    • we want people to be able to participate
    • participants should be reachable to participate
  • Our requirements / ideas
    • share data
      • logs
      • metrics port data
      • netflow data
    • we need realistic traffic
      • tgen / shadow may be able to help
    • we should not be useful as a general purpose tor network
      • existing tor exit operators can run exits
      • we can restrict exit to a “safe” set of sites and services
      • we should setup reachable locations with random files to download for testing
      • maybe directory authorities in the test network should not assign exit flags
    • should we have a chaos monkey-like thing?
  • there is chutney
  • adding new directory authorities to an existing set of directory authorities is apparent a process
    • it is best to introduce a new authority, keep it offline so it does not participate in voting initially

Actions:

  • alex talks with Hiro
  • scm wants to be in on this
  • generally community members are time-restricted. we aim for a best effort