As it is not installed on the target machine and it only generates error
messages.

Facebook uses some more assignments and spam the analysis by them.
Whitelist them as well.

Because pythons GIL ☺

This reverts commit 410107dd.

This reverts commit a80ea208.

This reverts commit 58697377.

So we push data faster. Maybe.

With smaller number of routers, try to fix some errors.

Allow a different filter for blue turris and omnia. Will get reverted
after a short while.

Like, at all.

Since we use our own CA, we need wget to verify it against that. Include
the certificate and link to it.

We don't want to catch „attackers" coming from lan, so restrict the
magic and apply it only on connections coming from WAN. Ask the system
what the WAN actually is.

And yes, this expects somehow sane network configuration, but it should
be default anyway, so this catches 99% of cases.

Test and fix the AmIHacked rust processing. Also include few
improvements:
• Use different hashes in hash tables, to speed up a bit.
• Provide tab in the output instead of a space, as is required by
  to_db.pl
• Accept both compressed and uncompressed input files, as needed by
  repu_add

Testing needed.

Splitting into functions, scoped threads

In parallel across all the input files.

• Don't ask for OCSP ‒ this may produce more output which confuses the
  parser and generates wrong data.
• Don't send the empty line to the server. This caused the server to
  generate a Bad Request, which we don't want.

If one client sends broken data, don't kill the whole batch for the
other clients.

The Mirai botnet use some more interesting ports.

The old version of the NetAddr::IP on the server doesn't know the
function and we probably don't care about storing up to two strange IP
addresses.

Omit addresses that shouldn't be on the wide internet. This is things
like local-network addresses, multicasts, etc. These just make little
sense as sources of attacks and it usually means something else than
attacks.

Use the attack kind as the first index and the date as second. First,
the resulting JSONs are likely to be smaller, second, it makes sense to
do a graph of one kind of attack over time, but it makes little sense to
do a graph for single day comparing pears and apples.