Hi all,
below's what I did in June. Yes, there's still one more day left in June, but I don't expect to do anything useful tomorrow other than packing for Italy and pretending to have a half weekend.
Best, Karsten
Fixed a few problems with the metrics website: bridge usage statistics graphs were missing 85% of reported bridge statistics due to a truncated file containing raw observations (#6064); download times in Torperf graphs were consistently 1 second too high because of a trivial math bug (#6070); some direct usage graphs showing possible censorship events displayed upturns or downturns for every single data point (first part of #6170); direct usage graphs of countries with little usage displayed missing values where they should display zero users (second part of #6170).
Added a new graph to the metrics website that shows number of cloud bridges over time (#5629) [0]. Runa is still investigating the sudden drops in March and April, but it looks like they're related to a problem with updating cloud images.
Visualized Torperf phases from starting a download to 1) receiving the CONNECTED cell, 2) receiving the first byte, and 3) receiving the last byte of the response (#6120).
Looked more into improved bridge usage statistics (#5807) with little outcome so far. The directory request statistics reported by bridges still look very promising for estimating daily bridge users.
Tried to find reasons for missing extra-info descriptors (#3266). There are some interesting patterns, but it's still unclear what's going on.
Finished making the new combined Torperf data format the new default format that we provide on the metrics website (#3036) [1, 2].
Added support for parsing GetTor statistics files to metrics-lib and made the metrics website code use that library instead of implementing its own parser.
Discussed sanitizing pluggable transport lines in bridge extra-info descriptors with George and implemented the required code in metrics-db (#4957).
Deployed George's changes to the censorship detector script to send out daily mails for censorship events [3]. Apparently, there are still plenty of false positives on that list, but the newly gained awareness might help improving the censorship detector script in the future.
Fixed a problem in DocTor, the consensus-health checker, which looked at old consensuses for some checks when it shouldn't.
Helped prepare the metrics host for a downtime of a few hours for getting its RAID card replaced. Looks like replacing the broken card has solved previously existing performance problems quite well.
Made a few smaller extensions to Onionoo, the data provider behind Atlas: added bridge nicknames; added support for searching by country code (#5960); included exit addresses in summary and search (#5251); added support for relays and bridges with multiple OR addresses (#5245); allowed searches for more than 1 search term (#5248); added reverse DNS lookup results (#5247). Also wrote a design document for Onionoo [4] to facilitate writing a Python version of it.
[0] https://metrics.torproject.org/network.html#cloudbridges
[1] https://metrics.torproject.org/data.html#performance
[2] https://metrics.torproject.org/formats.html#torperf
[3] https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-censorship-events
[4] https://gitweb.torproject.org/onionoo.git/blob/HEAD:/DESIGN