@SmokeDetector @PeterJ really? this looks like the OP attempted to fix the plagiarism problem by providing proper attribution (though of course still not a good answer per se)
[ SmokeDetector | MS ] Bad ns for domain in body, bad keyword in body, bad keyword in title, blacklisted website in body, link at beginning of body, +4 more (492): Weight Loss Pills 2020 by Cathy SSisto on askubuntu.com
@Magisch An autoflag of yours was used on this post [MS]. It was marked as NAA by some reviewers. You may want to take a look at the post and make your own evaluation.
[ SmokeDetector | MS ] Few unique characters in body, no whitespace in body, numbers-only title, potentially bad keyword in body (252): 111111111111111 ✏️ by pandaknight on stackoverflow.com
@Daniil It's stated in the PR that it's currently not tested against GitHub. It should be tested against GitHub prior to being merged. My expectation is that I'll get that set up in the near future. To do so, I wanted to bring up a new server. Doing that has been a bit delayed, but my expectation is that will happen today.
@tripleee Lumping everything in that tag as a single campaign is significantly broader than what I'd consider a "campaign" for the purposes of our "ongoing campaign" definition. IMO, to consider something part of an "ongoing campaign", there needs to be some tangible connection to a series of prior posts, not just that the posts are about the same topic area.
@tripleee Just FYI: There were no changes in the detection when it was moved to it's own reason. There are planned changes to reduce FP, but those haven't happened yet.
In addition, there is a very similar detection which remained in the watchlist. The one that remained in the watchlist is for YouTube only. It was a separate watch, and remained there, because when reviewing the initial results of the first overall watch it had a much lower %TP, but was the only thing that caught a significant number of spam posts. In other words, you'll still see some watchlist detections which appear similar to what the new detection will catch.
@user12986714 For what you appear to be doing, please use the GraphQL API. Doing so will allow you to make 1 API call for up to 100 posts, rather than the 101 API calls which will be required by using the feedbacks endpoint.
@user12986714 Just to be clear: If you're wanting to have all, or even just most, posts, then you should download the database backup and connect to a database on your machine running that. If you decline to do so and insist on loading them from the API, then you need to space those requests out over time, so they don't impact anyone else using MS. That basically means that using the API to download all posts should take you at least many hours, if not days to download all posts.
@user12986714 OK. I mostly just wanted to make sure that it was understood that my comments did not mean I was contradicting what @ArtOfCode was saying about using the database. (i.e. I should have read the entire conversation prior to posting my first message in this conversation.)
@user12986714 Yes, it's inconvenient. There was a recent comment on the "it's broken" issue for library we use for GraphiQL suggesting a workaround, so I'll see if implementing that works. Alternately, I have a request I'm planning to use with FIRE and AIM, which I constructed while the GraphiQL interface was working. I can give you a copy of the code for that.
@SmokeDetector I custom mod-flagged, because I've seen a lot of moderators just close this type of spam, when it really should be deleted, even if it's just a straight deletion, rather than a deletion as spam.
@Rob @Undo It looks like the logs for everyone's flags are no longer showing up. Here's SmokeDetector's. It is suspiciously close to when I hit the button to delete Shree's account.
I still see my last days worth (as do others), it would be great if "restoring" merged old records with newly created rather than a simple overwrite - but we *know* that. 🏃♂️
@SmokeDetector The watchlist version of the detection doesn't capture the URL in order to allow other watchlist entries to trigger on the domain. Now that it's its own detection, we should be capturing the URL, so people can make a better decision based on the why data.
@Machavity on-topic answers are still spam if they lack disclosure. If they are off-topic, they are spam even with disclosure (e.g. "You will like our new dietary supplements, click here to order this is my shop....").
That's the real problem. 4 self-answered questions that look like tutorials. Not sure how kosher that is, but I'm willing to bet the mods leave them without links
I think that's a good case for a mod flag, as Machavity suggested, because destroying the answers that contain useful non-link content is bad, but the user needs to be informed to knock it off.
FYI, here's the text of my mod flag: "Possible serial spammer? Each of this user's posts are questions that have been self-answered (some now as community wikis), and they ALL have/had links to his own YouTube blog(s) (some have now been edited out by others) ... this seems like abuse of the SO website. Sorry if I'm overreacting!"
