Skip to content

What replaces tor control-port + stem in arti?

session run by geko, notes by ln5

  • in c-tor there is a control port where you can ask tor for info and make it do things like building circuits
  • stem is a python client library for this
  • stem has been unmaintained for a while, partly because we have arti coming up
  • arti is not exposing a control port, so how will we be doing what our current python tooling is doing today?
  • current use
    • sbws is an example of a larger program
    • smaller scripts for all sort of things, f.ex. gathering numbers from a running relay for debugging tor
  • can arti-client replace stem?
    • not everyone want to keep a rust toolchain around for running a four line python script
  • background: wanted a better protocol than control-port, we want bidirectional comms over an rpc channel
  • current status: there is an rpc protocol, with python examples
    • textual protocol with a data type schema
  • currently ongoing: finish what TB needs
  • requested: onion servers management, f.ex. ricochet-refresh and cwtch
  • open question: support for what onion balance would need or not?
  • what are you using on the control port today?
  • formats for data on disk: like server descriptors; encoders and decoders, parse2
  • for operations like creating circuits
    • could we make a crate ("arti-analysis") compiling into a shared library with a python API?
    • target should be that such an arti-analysis can be imported into jupiter notebook
    • could also expose the document parsers
    • might want to try tot get a grant for this work
    • we could PoC this today/tomorrow
  • margot is a tool giving a rust api with a single method "run" giving a bootstrap directory
    • a "bootstrap directory" is what a client needs, like the consensus plus the microdescriptors
  • dbgor is another interesting tool, implementing something like a tor REPL
    • used a lot by the network team
  • could be part of what network-health team should be looking for funding for
  • rpc timeline: prioritizing according to a plan of sunsetting current usage of c-tor
    • arti relay within ~6 months