@Makyen When you say "performance", I think "speed", whereas I suppose you meant something closer to "accuracy" or "hit-rate". Nevertheless, quite good. Of course, I wonder if the percentage of TP is more a factor of its relatively recent introduction than a factor of its inherent accuracy.
Unfortunately, my Internet connection's performance is poor, which resulted in that message timing out when attempting to send, then getting posted twice anyway.
@CodyGray Unfortunately, this regex isn't going to do what you want, due to [\W_]*+\. never being able to match, because any "." will be consumed and not given back by the [\W_]*+. For what you appear to be wanting, I'd suggest r[\W_]*+b[\W_]*+g[\W_]*+y, or r[\W_]*+b[\W_]*\.[\W_]*+g[\W_]*+y, or r[\W_]*+b[\W_]*?\.[\W_]*+g[\W_]*+y.
@CodyGray Thanks for mentioning it. Sorry for the confusion. Changed to accuracy. cc @Spevacus
I've been thinking that it might be reasonable to generally relax the requirements for blacklisting phone numbers. It appears that even the watched numbers have high accuracy. The ones which we have seen a lot of FP for were numbers which were primarily or completely sequential or repeating. We could keep more stringent requirements for numbers which are somewhat like those, while having looser requirements for most phone numbers.
I agree. I've been tending to eagerly blacklist phone numbers that I think are almost certain to produce only TPs. tripleee has called me on it once, but I still think it's a reasonable tactic. I'm glad to have some empirical evidence backing up my intuition. :-)
@CodyGray no objections, but we should update the guidance in the wiki if we all agree
there are several things which should be changed, I have been meaning to draft some sort of update but I'm wondering whether this would be suitable for chat, Teams, the wiki, a Github issue, or something else
in particular, the "last two years" limit for blacklisting needs to be relaxed; I have been ignoring it in practice, though it would be good to have some limit
@tripleee If you're thinking more substantial changes, then it should be discussed more formally.
As for the number watch/blacklist, it's been my intent/expectation from the beginning that we would likely have different requirements for blacklisting numbers, but that we needed actual data from a substantial period of time as to the accuracy of both those we blacklist and those we've watched in order to get a feel for formulating what it should be.
In my opinion: A GitHub issue is best if the discussion is expected to be minimal (i.e., if we're not going to need to go back and forth to hash out a lot of details). If you are basically making a proposal, and looking for agreement or criticism, then a GitHub issue is fine, even ideal. If you're wanting to have a discussion, then chat is better.
(I feel like that is somewhat incoherent. I blame the fact that I've been writing JS for the past several hours. It always turns my brain to mush.)
@tripleee We can, or just put something into the wiki and go back and forth a bit. It depends on what other things you're wanting to change and how substantial that is. For just the phone numbers, I'd assumed I'd get general agreement that they should be looser, then pop a first version on the wiki and people could comment in chat and/or change it.
@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 word(s) if you really want to add the pattern you provided.
@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 word(s) if you really want to add the pattern you provided.
[ SmokeDetector | MS ] Potentially bad ip for hostname in body, potentially bad keyword in body, potentially bad keyword in title, potentially bad keyword in username, blacklisted user (74): Industrietore aus dem Ruhrgebiet by industrietor online on meta.SE (@Ollie @Ethan)
@SmokeDetector tp- undisclosed affiliation (email in about section of youtube channel linked from webpage matches username, user's linkedin page shows same physical address as on the linked webpage).
@SmokeDetector tpu- undisclosed affiliation. All 3 hits for the linked website on the network were posted by this user, and are recent. Custom flagged.
