@Makyen Yep, wording could be better.

[ SmokeDetector | MS ] Offensive answer detected (79): C# Double buffering causes lag on paint by Mikael Rindmyr on stackoverflow.com

fp- by Makyen

[ SmokeDetector | MS ] No whitespace in title (59): MultiValueDictKeyError by Maria on es.stackoverflow.com

fp- by K.Dᴀᴠɪs

[ SmokeDetector | MS ] Few unique characters in answer, repeating words in answer (173): Why are my leaves dissappearing and turning black by Nikhil Ghugari on blender.SE

tpu- by K.Dᴀᴠɪs

[ SmokeDetector | MS ] Few unique characters in body, no whitespace in body, repeating characters in body, repeating characters in title, title has only one unique char (420): aaaaaaaaaaaaaaaa ✏️ by Sesame on quant.SE

tp- by K.Dᴀᴠɪs

fp by Nisse Engström on Can't understand flac frame header format [MS]

@Glorfindel recall your thinking on why this was spam?

@Randal'Thor can you find out why this was deleted for me? Was it deleted for being NAA or as spam?

fp by Nisse Engström on Restricting access based on a field [MS]

fp by Nisse Engström on clearInterval SetInterval function not defined [MS]

naa by ArtOfCode on Use numpy.curve_fit to on a function with summation terms [MS]

fp by Nisse Engström on True random generation given radioactive entropy [MS]

@Undo - same on this one, was that deleted as NAA or spam?

@ArtOfCode What time point are you interested in?

October 30, 20:00-21:00

Rude for our purposes.

metasmoke.erwaysoftware.com/post/141471

@Undo nah, repairable, whatever the feedback for that is

f, apparently

but still NAA, so feedback would be NAA too

[ SmokeDetector | MS ] Potentially problematic ns configuration in body (27): add sidebar to a wordpress theme cartabox by Daniel Garcia' on stackoverflow.com

@Undo third opinion on this one? Looks like I said FP before, but I'm not sure now.

Looks kinda like Wikipedia

aye, could be plagiarised +spam

can't find the source

Finding references to http://oukyi.blogspot.com/2018/10/is-esd-serious-risk-on-modern-machines.html

dead blog, looks shady as heck

That's the one I found too

loads really slowly

Or it could be legit and just got scooped up by some domain grabber

... oh, it's an SU scraper

Post link points at a shady-looking Chinese cheapo-electronics retailer

huh. and a pop up made it past FF's popup blocker

That blog is spam

Calling that spam

burns down laptop for protection I'm so sorry machine. I'll be better next time.

oh, yep, yay macOS

and what about this post? What happened to that?

I'm on Ubuntu right now. So, probably good, but just to be safe

RemoveAbandonedClosed

closed as off topic by the community

Cool

fp- by Tetsuya Yamamoto

Guessing detected for that last line

!!/test newbecca.com

afk, eating time

Okay. Well in the last month, I can't find any posts that were conflicted+deleted and turned out to be not-spam.

Not sure if a month is a long enough time period to expect that to happen in.

tp by ArtOfCode on Experience with travelling long hours in a car with 10 week old baby? [MS]

tp by ArtOfCode on Experience with travelling long hours in a car with 10 week old baby? [MS]

tp by ArtOfCode on Experience with travelling long hours in a car with a 10-week-old baby? [MS]

[ SmokeDetector | MS ] Few unique characters in answer, repeating words in answer (173): Where do babies come from (on the Axiom)? by yolo cucumber on scifi.SE

tpu- by K.Dᴀᴠɪs

[ SmokeDetector | MS ] Username similar to website in answer (51): Load data from Google Analytics in a Chart.js chart by bryceandy on stackoverflow.com

So do y'all think it's worth me looking at doing auto-mod-flags?

fp- by Tetsuya Yamamoto

That is, if a post (a) has at least $threshold (start at 1, see how it goes) manual flags on it, and (b) has at least $otherThreshold (start at 1) FP feedbacks on it, and (c) gets deleted, cast a mod flag on it

That flag probably says something like "This post was flagged N times through Charcoal systems and has since been deleted, but was later flagged up as a potential false positive. Please review this post and make sure deletion was appropriate for it; let us know if it wasn't. If you're wondering WTF this flag is, see <some link to website>".

@ArtOfCode It should be pretty simple; the biggest question (other than "do we want it") is what account to use

Could build up another account for metasmoke that we don't use to cast spam flags, only 'other' flags

I'd use Smokey's, it's more likely to be recognised

@ArtOfCode ~4000 posts (about a month) is ~3% of our total volume. Not sure how the stats work out on it

@ArtOfCode Except we already cast the first flag in autoflag cases with that account

orrrr nevermind, spam flags are cleared by the time it's deleted

@Undo Should still be able to other-flag, no?

though that raises the question of whether the API will let us flag a deleted post

I think it will

IIRC I've seen the mobile app show only the 'other' box in cases where the post was deleted after page load

You've got the tokens; wanna test it?

I would but I misplaced Postman on the last laptop move

I could Keybase you your token

[ SmokeDetector | MS ] Few unique characters in answer (86): How to add a prefix to my domain name? by Steve on webmasters.SE

tpu- by Tetsuya Yamamoto

@Undo go for it

sent, confirm receipt here

got

Great

your erwaysoftware.com proofs of ownership are invalid, btw

bleh

DNS there is borked on another thing too

Disturbingly easy to extract tokens from MS w/ command line access

Not unexpected, but I'd still like to move 'em to another system at some point. That'd be a fun project if I had time

That's the thing about being the system owner :) It'll always be as easy for you as it is for the code

It'd be kinda fun to make a trapdoor system - tokens go in, tokens don't come out. Then the system just knows a few commands - flag this as this, invalidate token, etc.

If the code can do it, so can you - the Rails console gives you exactly the same environment as the code.

Code somewhere can do it, but that doesn't mean I need to have a console into it

TL;DR: If the tokens ever touch Rails, it's cake for you to get 'em

If you want to not have access, they have to not touch Rails

That'd be the project

Tokens pass straight through MS, never get stored

There are also architectures where tokens never hit MS, just go straight from SE to Knox

[ SmokeDetector | MS ] Potentially bad ns for domain in answer (1): org-mode: any way to automatically collapse "blocks" (e.g., BEGIN_QUOTE)? by Grant Shangreaux on emacs.SE

fp- by Tetsuya Yamamoto

@ArtOfCode This will still raise such a flag in situations where we internally reverse direction. For example, where there are some initial tpu- feedbacks and manual flags recorded, but where we decide, after discussion, that it's actually FP or NAA and people retract their flags, but that it still gets deleted as either red-flagged or NAA. Basically, MS has no visibility into users retracting their flags. However, it's not unreasonable to kick it to a moderator, even under those circumstances.

My expectation is that the total number of situations is low.

Another alternative would be to raise a report that an MS admin would need to handle, rather than sending it to site moderators.

If raising an info flag for site moderators, we should consider if we want to including both a links to the MS post and a link to the SD report in CHQ, where there may be more discussion than is recorded in MS comments.

What if feedback had to be conflicted at the time of deletion to trigger the flag, not just FP?

@ArtOfCode I have not actually tried flagging, but the API returns a valid option_id for the custom flag on deleted posts for posts on sites both where I have 10k+ and where I don't. Thus, I expect that the SE API will permit us to send custom flags on deleted posts.

[ SmokeDetector | MS ] Bad keyword in answer, toxic answer detected (168): Why and how were the Mongols able to invade Russia successfully? by mike hunt on history.SE

tpu- by Tetsuya Yamamoto

@Undo if I flag something on SO, are you around to catch it?

it'll be this post

yep

and can you keybase me MS's API key as well - can't flag anything with that token without the corresponding key

oh yeah

@ArtOfCode DcPBvNNjjUvD2xax9hNHpQ((

got

That's not secret, it's probably in Git logs somewhere anyway

okay, #1 flagged, that was a live post and should have succeeded

yep

Can you throw me the ID for a deleted post now?

@ArtOfCode stackoverflow.com/a/53291874/1849664

ta

wow. mod abuse. marking helpful a spurious flag. im going to tell on you

and flagged

So would this only check if the post was deleted on feedback invalidation

@quartata the custom message was great

@quartata not on invalidation, on post deletion

on post deletion, if post has FP feedback and manual flags, cast mod flag

or if post was conflicted, not decided on that one

it requires the post being deleted in a fairly short time frame

2 hours about

@ArtOfCode Handled, should make quartata happy

heh

thanos flag

Probably could have chosen a post from someone other than a 100k user, but hey.

@ArtOfCode Got a link to the thing you were using earlier for feedback?

Just search?

Couple of custom queries

metasmoke.erwaysoftware.com/data/sql/queries/…

metasmoke.erwaysoftware.com/data/sql/queries/…

there were more in that list, but I cleared the conflicts

and none of 'em had flags

!!/test bekausa.com

tp by Tetsuya Yamamoto on Java 8 streams, lambdas [MS]

tp by Tetsuya Yamamoto on Java 8 Lambda and Streams [MS]

tp by Tetsuya Yamamoto on Java stream operation invocations [MS]

@ArtOfCode So... Does all of this strike you as highly improbable?

hm?

Like, we can't be this good that we pass every audit we do

There's gotta be an ugly failure in here somewhere

[ SmokeDetector | MS ] Body starts with title and ends in url (97): How to sign on ethereum a pdf like on this video? ✏️ by Damian Joniec on ethereum.SE

@Undo somewhere, yeah. I can accept that it's relatively unlikely, but I'd expect to see a few.

So either we're really really good or there's something we're missing

Both are possible; the former seems most plausible at this point

[ SmokeDetector | MS ] Bad keyword in answer, blacklisted username (178): HEVC format convert to MP4 in ubuntu by daveclark966 on askubuntu.com

Given that we've come close on a few occasions, there's probably a very low number where something has gotten through.

MS drops posts occasionally when it does, but that should be independent of ugly failures

tpu- by Tetsuya Yamamoto

@Makyen I wonder if there's a weird distribution of stuff around that borderline. Perhaps the 'borderline' stuff we see gets caught for some statistically significant reason.

So our average invalidated feedback rate is around 5%, accounting for who's here when. Using a SW-AG estimate, I'd guess that somewhere around 5 people see each Smokey report. That's already 3.12e-7% chance of something being missed, without accounting for review or admin invalidations.

Complex subject

@ArtOfCode Not independent, but a starting place

And yeah, numbers are not the best way to approach that particular problem, but it lends some plausibility to "the system works"

@Undo Could be.

@ArtOfCode I suspect this is much lower than 5 during some times of day. Most people don't go back through the chat history. For many/most people, SD reports are hidden upon post deletion. For a significant period of time, we required only 1 feedback, which means things didn't, necessarily get reviewed.

That's why it's an SWAG estimate :P

[ SmokeDetector | MS ] Potentially bad ns for domain in answer (1): Absent Vs Missing by farhanfar on ell.SE

tpu- by Tetsuya Yamamoto

!!/watch- undangancinta\.com

[ SmokeDetector ] Blacklists reloaded at rev f85821a (SmokeDetector: Auto watch of undangancinta\.com by Tetsuya Yamamoto --autopull) (running on iBug/EC2)

@SmokeDetector Just in doubt: intentional spamming or testing waters?

@ArtOfCode That might be reasonable, although I'm not sure that I'd limit it to just FP/NAA when deleted (i.e. should probably set the flag even if we get that feedback after deletion).

aye

@TetsuyaYamamoto pernikahan in Indonesian means wedding.

@scaaahu I know it already, because I'm a fluent speaker of Indonesian and Malay.

[ SmokeDetector | MS ] Few unique characters in answer (86): Should my fighter really punch that werewolf? by skittles dude on rpg.SE

tpu- by Tetsuya Yamamoto

naa- by NobodyNada

[ SmokeDetector | MS ] Potentially bad keyword in answer (1): Why won't Israel just capture Palestine? by mugbean mclimmerick on politics.SE

@SmokeDetector n

tpu- by NobodyNada

@SmokeDetector Going with borderline tp -- unreg user, smells like holocaust denial, and calls someone "a failure"

[ SmokeDetector | MS ] Bad ns for domain in answer (72): Bonds with biological versus adopted children by Sequencefanatic on parenting.SE

[ SmokeDetector | MS ] Potentially bad ns for domain in answer, potentially bad keyword in answer (2): PHP array delete by value (not key) by Sunil on stackoverflow.com

@SmokeDetector 1 tp 6 month ago same user not sure.

fp feedback on autoflagged post: Unable to set the layout as expected using Bootstrap v4 alpha 6 MS

fp- by Makyen on Bonds with biological versus adopted children [MS]

@Undo Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@ArtOfCode Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@Ferrybig Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@Magisch Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

oh god

[ SmokeDetector ] SmokeDetector started at rev f85821a (SmokeDetector: Auto watch of undangancinta\.com by Tetsuya Yamamoto --autopull) (running on iBug/EC2, Python 3.6.6)

I DIDN'T SAY DISABLE THEM WHETHER THEY WERE VALID OR NOT

@ChrisBerger Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

goddammit

@ChrisBerger Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

Restart: API quota is 11334.

@River Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@River Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@Henders Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@Henders Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@Karelzarath Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@Karelzarath Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@FrenzyLi Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@FrenzyLi Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@TajyMany Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

oh ffs

@TajyMany Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@NVZ Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@NVZ Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@paper1111 Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

I assume the FP on autoflagged post was a test, not something we need to independently investigate. Correct?

@paper1111 Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

It was me deleting the FP feedback on that post -_-

@WELZ Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@WELZ Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@iBug Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

Hmmmm.... that's.... unexpected.

yeap

@iBug Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@Undo around?

@Zoe Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@Zoe Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@casualcoder Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@casualcoder Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@jhpratt Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@jhpratt Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@rollstuhlfahrer Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@rollstuhlfahrer Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

@xfix Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

bahahaha

@robinCTS Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

How'd you pull that one off?

!!/reboot

@robinCTS Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

not a clue

Reboot should drop the messages in the queue

(a) deleting the feedback really shouldn't have triggered an autoflag FP alert

[ SmokeDetector ] SmokeDetector started at rev f85821a (SmokeDetector: Auto watch of undangancinta\.com by Tetsuya Yamamoto --autopull) (running on iBug/EC2, Python 3.6.6)

Know which post it was?

(b) the query that figures overall accuracy shouldn't be flagging all of these as invalid; they're all over the threshold

metasmoke.erwaysoftware.com/post/142350

Restart: API quota is 11303.

I assume it disabled everyone's flag conditions that had one with weight set it slightly below the 280 default (i.e. at or below the 257 on that post). I'd be good to find out if it disabled those who only had conditions above 257.

fp by Nisse Engström on Onset of Autophagy [MS]

It didn't kill all of them

@Undo can you pull prod logs for ActionCable broadcasts of that "flag conditions were manually disabled" message in the last hour? I want a complete list of everyone who had conditions disabled so I can go back through and re-enable them

It's gonna be easier for me to do it

fp by Nisse Engström on how do I install Bitcoin Private electrum wallet on ubuntu 16.04 LTS? [MS]

Just going to pull a backup and move that one column across

@Undo heh, that also works

decent test of DR procedures at the same time :)

@Makyen Your flag conditions were manually disabled by @SmokeDetector: Total accuracy must be >= 99.75%.

and look at that, I even have backups!

>:/

Some job must still be running

le sigh

you might have to reboot MS, it's in a thread

yep

@SmokeDetector OK, that means it's still actively disabling them, as I'd re-enabled some conditions, which certainly were not below the threshold.

just gonna reboot the whole thing

Half a chance it doesn't work, though, there's a reason it's doing this

not a clue why this is happening, but I'm reverting the commits that touched this logic yesterday - it worked before then, if badly

while you're on the server anyway, #528 needs your attention

[ metasmoke ] ci/circleci success on 4f67f12: Your tests passed on CircleCI!

[ metasmoke-deploy ] deploy started by ArtOfCode

tpu- by Tetsuya Yamamoto

@SmokeDetector Seems this spam-flagged by mod, with -2 votes.

@ArtOfCode I just clobbered the whole table with this morning's backup. Should be close enough

works for me

If you happened to add/remove/change a flag condition in the last 12 hours or so, it's probably reverted. Sorry ;)

oh

@ArtOfCode what was github.com/Charcoal-SE/metasmoke/commit/… for?

usually you need a return to return stuff

yeah, that'd do it. No early return, so it disables everything it sees

seven char fix for that

[ SmokeDetector | MS ] Blacklisted website in body (99): How to fix install blog extension last step error? by neha rani on magento.SE

fp- by Tetsuya Yamamoto

[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, blacklisted website in body, link at end of body, pattern-matching product name in body, +4 more (692): What is Keto Ultra New Zealand? by user963482 on superuser.com

tpu- by K.Dᴀᴠɪs

@ArtOfCode Inclined to leave 'em as is, thoughts?

@Undo I'd pin 'em. Not a whole lot of effort, might prevent a post getting over-weighted and flagged when it shouldn't

That works too

At least one already had the automatic pin for too-little-data

tpu- by Shree on Can I place a comment inside a tag in ASP.NET? [MS]

Pinned the three 'potentially problematic' ones, the others look pretty mature

works for me

[ SmokeDetector | MS ] Potentially bad keyword in body (1): Importing csv to mysql and an array value to first column of table by Jeremy Adrian de Vera on stackoverflow.com

[ SmokeDetector | MS ] Url-only title, blacklisted website in body, blacklisted website in title, body starts with title and ends in url, link at end of body, +5 more (794): evaherbalist.com/beligra/ by Suwbpafy on askubuntu.com

tpu- by K.Dᴀᴠɪs

fp- by Tetsuya Yamamoto on Importing csv to mysql and an array value to first column of table [MS]

What do you make of this: OP linking to their own site for no apparent reason?

pretty common in "help me do this thing" questions

they have three previous posts which were deleted as spam metasmoke.erwaysoftware.com/domains/21788

[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, potentially bad keyword in body, potentially bad keyword in username (197): How Much Cost of Android App Development Course in Delhi Institute? by HTS SEO on security.SE

tpu- by K.Dᴀᴠɪs

!!/blacklist-website htsindia\.com

3/3 convincing enough

[ SmokeDetector ] Blacklists reloaded at rev de16d35 (SmokeDetector: Auto blacklist of htsindia\.com by tripleee --autopull) (running on iBug/EC2)

[ SmokeDetector | MS ] Url in title, bad keyword in body, blacklisted website in body, blacklisted website in title, potentially bad keyword in body, +1 more (394): amazonhealthstore.com/where-to-buy-keto-direct/ by sephheeler on graphicdesign.SE

tpu- by K.Dᴀᴠɪs

[ SmokeDetector | MS ] Potentially bad keyword in answer (1): Smart Thermostat Keeps Losing Power by getSurreal on diy.SE

fp- by K.Dᴀᴠɪs

[ SmokeDetector | MS ] No whitespace in title (58): MojoFailureException by Padawan on ru.stackoverflow.com

fp- by Tetsuya Yamamoto

@SmokeDetector ignore-

[ SmokeDetector | MS ] Potentially problematic ns configuration in answer (43): How to protect my digital data from Solar Flares? by user963504 on superuser.com

tpu- by Tetsuya Yamamoto

!!/watch- 365solar\.xyz

@TetsuyaYamamoto That pattern looks like it's already caught by Potentially problematic ns configuration in answer and Potentially problematic ns configuration in body; append -force if you really want to do that.

!!/watch-force- 365solar\.xyz

[ SmokeDetector ] Blacklists reloaded at rev f99b06e (SmokeDetector: Auto watch of 365solar\.xyz by Tetsuya Yamamoto --autopull) (running on iBug/EC2)

[ SmokeDetector | MS ] Username similar to website in answer (51): error messages when enabling /server-status on Centos with Apache version httpd-2.2.3-31 by Jamshaid Ali on serverfault.com

@SmokeDetector n

[ SmokeDetector | MS ] Username similar to website in answer (51): Forbidden (403), when accessing server-status on apache 2.2.22 by Jamshaid Ali on serverfault.com

tpu- by Tetsuya Yamamoto

What happened to Smokey?

happened when? seems to be alive

!!/coffee Zoe

@tripleee brews a cup of Cappuccino for @Zoe

Tonight. Saw the wall of flag condition invalidation

Oh

[ SmokeDetector | MS ] Potentially bad keyword in body (1): How do I determine what dialect to use for a Chinese translation? by Sandra Lee on chinese.SE

Whoops ^^"

tpu- by Tetsuya Yamamoto

!!/watch- tridindia\.com

@TetsuyaYamamoto 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.

@SmokeDetector k

Already in 3, should I wait other 2 for blacklisting?

[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, blacklisted website in body, pattern-matching product name in body, pattern-matching product name in title, +2 more (592): KetoFit Canada KetoFit Canada by qlaanskad on apple.SE

@TetsuyaYamamoto I'd recommend it

tpu- by Tetsuya Yamamoto

!!/blacklist-website tridindia\.com

[ SmokeDetector ] Blacklists reloaded at rev f1c8449 (SmokeDetector: Auto blacklist of tridindia\.com by Tetsuya Yamamoto --autopull) (running on iBug/EC2)

(I meant I'd recommend waiting, but don't worry about it -- I don't think it really matters.)

Whoa, I misinterpreted your recommendation. My bad.

single account, all below auto; leave it blacklisted

@TetsuyaYamamoto not a problem at all; I should have been less ambiguous.

Also the site is unlikely to get any FP feedback.

agreed.

@ArtOfCode it's not spam, it's (borderline) trolling. I would definitely like to have a system level block on those kind of posts.

@Glorfindel is there context I'm missing? It looks like someone who thinks we're reddit

just noticed that someone changed the login button on chat to link to MSE login rather than the old openid

good enough

@NobodyNada does Reddit allow questions like that? Then I'm glad I never visit it.

I have no additional context BTW.

@Glorfindel I don't use Reddit either and I guess it would depend on the rules of the specific subreddit. I think that post looks like the sort of thing you'd find on a really lax programming messageboard.

They're not blatantly promoting anything (no link) or harassing anyone -- just being chatty

Meh, could be. Still, I want a system level block on those kind of posts.

So maybe a TP feedback is more appropriate than TPU.

@Glorfindel eeh, sort of. The "system level block" argument sometimes makes me cringe because it can be applied to anything (NAA for example).

IMO "system-level block" specifically means a SpamRam block for extreme cases.

Though, to be fair, this is probably the most borderline post I've seen.

Yeah, I can certainly see why one thinks that's a false positive. Even more when you consider it has been caught 'by accident' (holiday.it is not a website link here).

If we had a 'gray area' feedback type I would've probably used it.

My personal stance for this sort of case is "we want to catch spam, and this isn't really spam"

[ SmokeDetector | MS ] Url in title, blacklisted website in body, blacklisted website in title, body starts with title and ends in url, link at end of body, +5 more (791): evaherbalist.com/beligra/ by Qmhvgmvg on drupal.SE

"We weren't trying to catch this post, therefore it was an fp"

tpu- by NobodyNada

[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, body starts with title and ends in url, pattern-matching product name in body, pattern-matching product name in title, +1 more (589): V10 Plus Male Enhancement by LonnaFinke on wordpress.SE

tpu- by Shree

HeatDetector is an example of a project with a different goal: it wants to catch problems, not just arguments specifically. Therefore, anything problematic is a tp, even if a comment is pointing out spam or plagiarism, or a different comment than the one caught is argumentative.

But since our project is looking for something specific, our feedback should reflect that specific metric. For example, I wouldn't give Natty a tp because it caught a plagiarized post; that's outside of the scope of the project.

Similarly, IMO a SmokeDetector tp is only for when something is definitively spam or R/A, since blurring the lines dilutes the potency of a tp. Finding an egregious question may be helpful, but it's outside the scope of the project because we're spam-hunters, not problem-hunters.

Vandalism is a bit of a special case: we don't try to catch it, but if we find it we give it a tp. This is because Smokey cares about content, not posts (so it doesn't care what a post used to look like). Giving it an fp would act as if we told the system "this content isn't spam" rather than "this post isn't spam."

[ SmokeDetector | MS ] Bad keyword in body, bad keyword in title, blacklisted website in body, blacklisted website in title, link following arrow in title, +5 more (793): BUY>> supplementforus.com/keto-ultra-nz/ by aplozmaklaz on superuser.com

tpu- by NobodyNada

NAA is another special case: it means "this answer shouldn't have been caught, but it is egregiously bad." If we had NAA for questions, the post under discussion would have met that criteria. (However, an NAA-for-questions feedback would be really subjective and varying site policies would be difficult to keep track of.)

!!/scan stackoverflow.com/a/53294669/3476191

@NobodyNada Post 1: This does not look like spam