Any thoughts on this user? 19/29 answers link to their own blog. For the ones I've looked at, the links appear to be relevant, answers are not link only, and they always disclose affiliation. There's no record of them in MS.
There are so many answers that only marginally answer the question, are so overtly self-promotional, and in such a short period of time that I think it's over the line.
@RyanM That pattern looks like it's already caught by Blacklisted website in answer and Blacklisted website in body; append -force if you really want to do that.
@cigien Looks like a legitimate question. The site it leads to is spammy in nature, sure, but I'm sure they're legitimately working on it and actually had a problem with it.
Until they're actually going to spam their site, I don't think we need to watch it. Besides, if they were going to spam their site, and they're going to continue hiding behind cutt.ly, what good is watching the site it links to? Wouldn't we rather watch what they're likely to spam? (The shortener extension, tjuH6lJ in this case)
@Spevacus Oh, right of course. And you've made this point to me before :( Sorry, I'll try and remember not to watch things that wouldn't be caught within the text of the post.
Yeah this circles back to the idea of creating a "When should I watch something, and what are the current watchlist policies Charcoal uses?" canonical in Teams. I feel like there's a lot of implicit stuff that gets brought up to newcomers/people who have taken some time off and aren't up to date on how we like to watch certain domains or domains that are watched but under a tighter regex (fiverr.com comes to mind)
5
I'll see if I can't start writing something up in the near term. Others can, of course, add what I will inevitably forget to include.
That sounds great. Yeah, the fiverr.com one tripped me up as well. I notice Makyen's message about it has been starred heavily, but that's not really a sustainable way to keep track of this stuff :p
@Spevacus Making a do-not-watch list is potentially useful for domains that seem like watching them is a good idea until you see the slew of FPs. I fixed the fiverr case in particular (now triggers the "already watched" warning if you try), but that won't always be practical, especially for cases where the answer is "we actually don't want to watch this at all, even with a stricter regex"
In case it's not obvious because stars are anonymous: I also very much agree with your idea, though it might be even better if it's periodically copied over to charcoal-se.org
(Teams is much easier for editing given that there's no need for a PR and every contributor already has the required SE/MS accounts, but more visibility is good)
If you had any others you were familiar with, you could PR them as additions to those whitelists. I'm sure you have at least a couple others you've run into.
@RyanM If you do add some, then be sure to have searched and made sure that you're not eliminating TP.
However, I would prefer that you wait at least a while to do so, as I am working on a list based on what's been previously caught for those two reasons.
@Makyen In theory, the query should cover most of that (it counts TP), but I'm happy to let you take a first pass.
That said, I would not personally consider losing something like metasmoke.erwaysoftware.com/post/199268 and metasmoke.erwaysoftware.com/post/199266 removing TP (which would result from whitelisting caniuse.com), as the domain is not spam, and the fact that it was TP was pure coincidence and had nothing to do with the domain. It could just as easily been any non-spam domain that they decided to link.
(caniuse.com, for reference, is linked in over 17k SE posts, and I would in fact recommend we not detect posts solely because it's linked)
@RyanM While this looks like a decent start, it needs to be separated out by reason, not combined. It needs to not include any domains which have any TP, and it should have a minimum FP count, which we'll have to look at wrt. what's appropriate, but should definitely not be below 3, with 5 probably more appropriate.
Yeah, definitely not suggesting all of those be added: it was intended for human review of the top ones.
Out of curiosity though, is there a reason they need to be split into multiple whitelists in the code? If we've reviewed a domain and determined it isn't spammy, does it matter which reason it's caught for?
(in an ideal world, I think domain tagging would be great for this, but we'd need some serious cleanup of the current tags to make it viable)
Or to put it another way, would we ever want to whitelist a domain for one IP/NS-based detection while still allowing it to be caught by another IP/NS-based detection?
@RyanM Currently, domain tagging doesn't work for this, because there are too many people doing just whatever they feel like with respect to tagging, without regard to what the actual effects are. While it might be possible to have a set of whitelisting tags, what we have now is a failure.
If it'd be helpful, I would be amenable to helping clean some of those tags out, but I'm currently reluctant to expend time doing so given the issues with the system that mean we don't want to increase the usage of it for detections. I'm also fine submitting PRs for updates to the list for the time being, and seeing if that becomes burdensome enough that something more dynamic ends up being needed.
A thought: One issue with the tags is the lack of visibility of changes: changes to the lists (except for those done with direct write access) all come through the channel one way or another, and we give each other helpful feedback on how things could be done differently. There's no such visibility for tagging, so mistakes go unnoticed and uncorrected.
I dug, and dug, and dug through github's Blame for when that site was added to the blacklist, but I could not for the life of me find the commit that added it.
There are 3 FPs (one dupe report, so 2 technically) on MS, and 2 standing FPs on an SE search.
LOL I saw your comment, I was surprised that they were like "Alright... I'll just trim that out. Can't get rid of the whole answer tho, gotta make sure my spam sticks around!"
@Yatin That pattern looks like it's already caught by Potentially bad keyword in body and Potentially bad keyword in answer; append -force if you really want to do that.
The Tavern's alright with it provided that Smokey has it and it's been alive for over a few hours, though it's generally best that it's been alive for longer than that. The Tavern's been dead for the past few hours, so I figure whoever's there wouldn't mind the additional message.
This user has posted 9 answers with links to their GitHub project without disclosure in the last hour. What is the appropriate response other than flagging the posts? SD watch?
Make that 10
You'd think the system would be able to stop that one.
@IanCampbell It depends on what they are doing. In this case, it would be reasonable to blacklist the user. What they were doing was well past reasonable and definitely into just promotion/spam. Usually, we want to educate users so that they comply with our requirements, particularly users with other contributions. So, ideally, the user would start/return to making beneficial contributions. But, until that actually happens, it's reasonable to watch and blacklist them.
Basically, 10 posts in 18 minutes promoting anything is way beyond what's acceptable.
[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, pattern-matching product name in body, potentially bad keyword in body (292): Don Lemon Cbd Oil by JeryThom on apple.SE
