Discobot tutorial flagged as spam

[ryantm (Ryan Mulligan)]

I am administrator on https://discourse.nixos.org/ which is hosted for free by Discourse.

It looks like the system flagged someone’s Discobot tutorial posts:

This is a link to the private thread about that (maybe someone at Discourse support can look at it?) https://discourse.nixos.org/t/greetings/2531

Is there something we have misconfigured about the spam flagging that would cause this?

Also, whenever there is a post flagged by the system, we have to process the flag and also the moderator inbox message. Is there a way to avoid having to do two things for every flag?

[tshenry (Taylor)]

Hi Ryan,

So, while I wasn’t able to fully reproduce this behavior, likely due to a spam-check job that runs at a certain time, here is what I believe happened:

• The site setting newuser_spam_host_threshold is defaulted to 3. The description of this setting is:

  How many times a new user can post a link to the same host within their newuser_spam_host_threshold posts before being considered spam.

  When digging into the code a little, it looks like this does not apply to PMs, so it does not appear that the tutorial itself would trigger the flagging.

• At 9:41am the System user flagged 4 posts at once for the user in question. Each flag had the same message:

  This new user tried to create multiple posts with links to the same domain (<link>). See the newuser_spam_host_threshold site setting.

  Between the first two flagged posts (non-tutorial posts), a link to the same host appears 3 times, thus hitting the threshold.

• The other two flags listed are from the discobot tutorial, but are referring to the naturally posted links (wikipedia and image) and have not met the threshold under this user.

• Main Suspicion: When a user breaks the spam host threshold, ALL posts (including PMs) where they have included links are included to make sure they are dealt with if they are confirmed a spammer.

So technically speaking, I think it’s doing what was intended. It looks like you’ve configured white_listed_spam_host_domains with some common hosts, but if you keep getting false-positives, you could try raising the newuser_spam_host_threshold sightly and bring it back down if there are any issues.

Let me know if you need any more information.

[ryantm (Ryan Mulligan)]

Your main suspicion sounds correct to me. Thanks for pointing out those settings, I just discovered white_listed_spam_host_domains today, so setting that should help a lot!