GitLab

TCP scanner for DNS flag day 2020

See merge request !14

Tools later in the pipeline now require argument specifying
type of scan (= set of tests) which will be performed.

Names of tools were generalized and some names of output files were
changed, see updated docs.

FIXME: This must be fixed before release for public.

Some of TCP errors were errorneously clasified as "compatible" but they
in fact cause resolution breakage.

WARNING!
Results obtained by previous version of tcpscanner are invalid and have
to be re-evaluated by executing tcpcomp2pickle and ednsevalzone again.

Idea behind this is to re-use preprocesing for TCP testing etc.

randomize order of queries in domain2ipset.py

See merge request !13

A strict RRL configuration might skew results, let's try to avoid that
by sending genreport queries in sequence. Given the large number of IP
addresses we test it should not considerably slow down the test.

Randomization should help with timeouts and response rate limiting.

Use IPAddress objects instead of strings

See merge request !12

It turns out that genreport and dnspython use different string formats
for IPv6 addresses and this is causing problems in certain special
cases.

Thanks to Sven Van Dyck <svenvd@dnsbelgium.be> for reporting this!

docs: fix Docker examples

See merge request !11

Dockerfile: optimize dnspython using Cython

See merge request !10

Docs overhaul

See merge request !9

rework result printing

See merge request !8

new script printresults.py replaces old nsprint.py and diffresults.py

Bugfixes

See merge request !7

The multiprocessing module by default uses fork() on Linux
and this effectively doubles amount of memory required for processing
big zones.

We are targeting Linux platform anyway so using forkserver should not
have any adverse effect.

This algorithm works better on large zones and produces the same result.

All-in-one script

See merge request !6