The problem I see is that without hellbans (putting someone in their own isolated environment where they think they are accomplishing what they want, posting spam, but really they are the only ones that see it), any direct restrictions will likely just be worked around, if the spammer is persistent, particularly when the spammer gets immediate feedback that they can't post something.
@Makyen how about scrolling down the info page to get the informed badge? can a bot emulate that (I'm asking since i never wrote a bot)? If not, that could be used as a cure... No more questions without the badge
@Vickel Sure, just a bit more work. Actually not much, as it would just be an appropriately times AJAX call to tell the server that it had been done, even if it hadn't.
@Spevacus Yes, all of those are common terms for it. They refer to the same effect. Jeff Atwood used to call them "hell-bans", and famously promised that we would never do that, so that's why this term is probably used most frequently here.
@cairdcoinheringaahing It actually refers to a very specific thing, not just spam in general, as the name might suggest. See also: meta.stackexchange.com/questions/332475/…
@tripleee That pattern looks like it's already caught by Potentially bad keyword in body and Potentially bad keyword in answer Append -force to the command if you really want to add this pattern.
[ SmokeDetector | MS ] Body starts with title and ends in url, link at beginning of body, potentially bad ns for domain in body, potentially bad keyword in body, repeated url at end of long post (238): Happy Birthday Images HD by happybirthdayimagehd on ai.SE
[ SmokeDetector | MS ] Link at beginning of answer, potentially bad ns for domain in answer, potentially bad keyword in answer, username similar to website in answer (103): Java to replace fields in HTML by Giftechies on stackoverflow.com
@tripleee That pattern looks like it's already caught by Potentially bad keyword in body and Potentially bad keyword in answer Append -force to the command if you really want to add this pattern.
occasionally it would be useful to be able to distinguish between "watch just in case" and "watch because this is spam but we don't yet have enough to blacklist"
the meta-question is "how complicated can we make the system before it becomes impossible to keep in your head" but I have been wishing for expiring watches too
Having some automatic component to it, like expiring after a certain period of time or "experimental" watches that expire after some number of FPs, would make it a lot easier to keep in your head.
Just adding another watchlist type/category is not necessarily valuable, because it increases the complexity and puts the burden on the operator. But adding some automation has less risk of that.
Thinking about it a bit more, I think I'd definitely most like to see a sort of "experimental" watchlist that will automatically prune things after a certain number of FPs or failing to achieve a certain ratio of TPs to FPs after a certain minimal number of hits.
we have discussed putting in a threshold but as the system matures, we have seen again and again that stuff that we watched a long time ago occasionally comes back
Yeah, but I wonder what significance that has. In other words, if we watched everything on the Internet, some portion of it would eventually show up as TPs.
one thing which Halflife could do would be to keep track of discarded watches and activate them again if a matching post gets a TP, or something like that
provided I suppose it never had any FP hits, as an additional condition
@CodyGray yeah, I have a thing which listens on the metasmoke socket and performs additional analysis on the reported posts; the results are displayed in the Charcoal Test room
yeah, I have a couple of branches where I tried to rehash the architecture but I bumped into snags with things like metasmoke's GraphQL API which I wanted to start using but which turned out to not support all the operations I need
(probably a bug, but far outside my competence to diagnose, as I know neither Ruby nor GraphQL)
@tripleee An invalid pattern was provided, please check your command.
@tripleee That pattern looks like it's already caught by Mostly punctuation marks in body and Mostly punctuation marks in answer Append -force to the command if you really want to add this pattern.
[ SmokeDetector | MS ] Bad keyword in answer, link at beginning of answer, potentially bad asn for hostname in answer, potentially bad ns for domain in answer (123): SMS Gateway Suggestions Needed by Arun sankar on serverfault.com
[ SmokeDetector | MS ] Pattern-matching website in answer, potentially bad asn for hostname in answer, potentially bad ns for domain in answer, potentially bad keyword in answer, username similar to website in answer, +1 more (218): Can I make a hot mayonnaise? by Mayola Pakistan on cooking.SE
@CodyGray I may have misled you: I made a mistake when writing that. It should have been (?-i:QUICKBOOKS)(?="(?<="QUICKBOOKS")) (note the " at the start of the positive lookbehind)
