[tor-dev] Interested in GSoC opportunity

Philipp Winter phw at nymity.ch
Mon Mar 14 17:30:16 UTC 2016

On Mon, Mar 14, 2016 at 12:06:20PM +0530, Deepankar Tyagi wrote:
> Extended goals:
> #1 create a module which does anomaly detection(machine learning)

Unless you have a very specific plan, I would advise against this.
Sound applications of machine learning are time-consuming, and Tor's
setting is particularly difficult to tackle as a large fraction of a
relay's information (e.g., nicknames, port numbers, bandwidth values)
are controlled by an adversary.

> #2 write behavioral and integration tests.

Sounds great.  I would also really like exitmap to be
Python 3-compatible.

> #1 In bin/exitmap "#!/usr/bin/env python" is used instead of specifying
> python2 or python3, Can I assume that code currently contains nothing which
> makes it biased towards one python version or the other. If yes then should
> one keep python2 in mind when developing ?

For now, exitmap only supports Python 2 although there's probably not
much that prevents it from being Python 3-compatible.

This assumption should be made explicit, so I replaced all
"#!/usr/bin/env python" lines with "#!/usr/bin/env python2" in commit
a4dccea.  Thanks for pointing this out.

> #2 As per the list subscribers' opinion, should the unit test be written
> after implementing core functionality only or is it better to write test
> alongside after each iteration of finishing parts of task? (in context of
> exitmap).

Personally, I don't feel strongly either way as long as they are

> #3 Should one aspire for as much less dependency on external libraries
> (other than stem) or if no then should one follow a rubrik when selecting a
> library ?

I think we can decide case by case if we want to rely on a new library.


More information about the tor-dev mailing list