codidact
Nerf posts that have been flagged as spam
Related to meta:291174
The linked meta post requests a spam reaction to alert readers, but that risks people reacting but not flagging (and then the post never gets addressed). Also, reactions are not anonymous and maybe this should be. An answer suggests disabling links if a post has a spam flag. Let's build on that.
Proposal: if a post has a threshold number of spam flags, disable links. Because that could be confusing, also add a notice along the lines of "links have been removed pending human review" to the bottom of the post. Spam isn't the only problem case here (there's also malware), so I'd rather not say "spam" in the notice, but we can certainly wordsmith this.
As for the threshold: a single flag seems consistent with close votes and our current small size, but I don't know if that will always be true, so let's code it in a way that's easy to change later, or maybe even make it a site setting.
Doing this would make meta:291184 redundant, I believe.
The easy way to do this would be a JS flag that sets links' href to javascript:void(0) if the post has been flagged as spam. That wouldn't neutralise links for users without JS, though.
The harder but proper way to do it would be to trigger an automatic edit (including post history event) when a spam flag is created to edit links out of the post.
The easy way sounds reasonable to me. Plus, if we do it the harder but proper way, we also need to catch if those flags are declined and update the edit history again. A runtime check for "is there currently a spam flag?" seems more resilient. (Also, we wouldn't want to attribute those edits, so that's more complexity.)
As I’ve mentioned before (I don’t really agree with this whole idea, but I’m outnumbered there), this does let flaggers create far too much noise when they, in fact, only have the ability to flag and not handle flags, for a reason. Automatically editing because of flags gives flaggers quite a big opportunity to cause permanent massive disruption and noise. If CD had a post timeline like on SE, you could at least minimize the impact by putting it there instead, if flags are going to change post properties. Automatically editing is just a bad idea.
Can't we just derender links for posts with spam flags as it's trivial to get that info via a query? Will require a couple of changes to the views and the controller, but worth it in the long run, imo.
This way we also have the benefit of, if users are fast enough, preventing those links from being forever embedded into scraper sites and the IA.
Can't we just derender links for posts with spam flags as it's trivial to get that info via a query? Will require a couple of changes to the views and the controller, but worth it in the long run, imo.
This way we also have the benefit of, if users are fast enough, preventing those links from being forever embedded into scraper sites and the IA.
That's a far, far better idea, although I'm still skeptical (however, that skepticism can rest on Meta). It still lets bad flags cause frustration and confusion. Legitimate users may end up asking on Meta why links in their posts aren't working. Having a question for that on Meta isn't an issue, but it puts extra toll on the victim. It might also lead to comments under posts, but those will primarily be of permanent disruption to the post author, curators and moderators, however, that disruption will be minimal, and is more of a fault with the comment system as it stands now.
So in terms of actually letting flags have an immediate effect, this seems the best possible one.
Also, if this is implemented, it may be worth looking into doing it for R/A flags, too, as posts can link to blatantly offensive or abuse websites without in itself being spam. From experience on SE sites, I think it would be wise to not activate it for R/A flags in the meta category.
How much toll will this put on curators and moderators, if the link is disabled?
If it's a disabled titled ([title](mylink.com)) link, there will be no way to easily grab the link without crawling through the post source, manually marking and copying the link out of it.
Looks like I filed two issues here (oops) -- foiled by my "clever" use of an answer post ID on the other one instead of the question ID here.
Duplicate of https://github.com/codidact/qpixel/issues/1333 ; there's discussion on both.
Is it possible to derender the link if there is no logged-in user but leave it for those with accounts? That way users can see it to add flags, curators and moderators can see it to evaluate those flags, but the link won't be there for everyone else on the Internet, which should greatly reduce the benefit of spamming.
We've been getting a lot more spam attacks in recent weeks, so I'm bumping up the priority of this.