Differences with sbws

1. External API:

   1.1. Bandwidth File

   • onbasca takes the observed bandwidth mean instead of the last one (solves tpo/network-health/sbws#40083).

   • onbasca scales the bandwidth using the last descriptor, not the one available while measuring (solves tpo/network-health/sbws#40081).

   1.2. Scanner

   • onbasca only measures relays with a descriptor (solves tpo/network-health/sbws#40076).

   • onbasca does not try to download the data 3 times adjusting the size according to the time it took, as sbws does (see #2, #3).

     In the future it should use a better algorithm than sbws (see tpo/network-health/sbws#30232, tpo/network-health/sbws#29291).

   • onbasca does not try to measure as entry an exit that fails to be measured as exit (solved in sbws by #40041 and others). Maybe this should be re-implemented.

   1.3. Logs

   • onbasca rotates logs by days, not bytes, as sbws does (solves tpo/network-health/sbws#40066).

2. Behaviour and/or trend days after an incident:

   2.1. Web Server

   • onbasca does not enable/disable the Web Server (#18).

     In the future it should use a better algorithm than sbws, cause it was prone to bugs (see tpo/network-health/sbws#40062, tpo/network-health/sbws#40067).

3. Tests

   • onbasca assumes congestion control enabled (cc_alg=2, bwscanner_cc>=1) while sbws only enable it for congestion control tests. The reason is due sbws needing to be compatible with previous tests.