« first day (2495 days earlier)      last day (1328 days later) » 

9:00 PM
tp feedback received
fp feedback received on [MS] I cannot find the image in this post
 
@user12986714 note that it was rejected for exactly the same reason:
> No, let's not do this. This just introduces potentially bad data. I'd rather see the post records deleted than potentially add erroneous feedback
 
@Daniil No, it's not. Right now, I don't trust that you won't move ahead with giving automated feedback. You've repeatedly said you're going to go ahead and do it, despite that we've, multiple times from multiple people, have said that we don't want automated feedback to be applied to MS. As you've also said, it's would be difficult for us to determine what is automated feedback and what is human feedback when either or both comes from your account.
Under those conditions, our only recourse is to not permit you to give feedback. The way that can be accomplished is to restrict your account to only the Flagger role. Thus, you now have only the Flagger role.
 
@NobodyNada So there comes the problem; if we can prove (with data) that human make more mistake than a machine, we can kinda argue that we would rather have posts deleted than have human review them because that 'generates bad data'
 
user435118
@Makyen So... how can I get the other roles back?
 
@user12986714 This is similar to how we accept that autoflags are accurate enough so that we can lower the requirements to only 2 manual reviews (flags) rather than 6.
 
I would really like if this whole discussion of using 1 feedback to reduce the queue can be discussed at stackoverflow.com/c/charcoal/questions/645 . I would like to participate as I have some professional expertise but I can't follow these short chat messages interjected with drama.
9
 
9:06 PM
@DanielWiddis Sounds good to me.
 
Unrelated to the topic, but missed in scroll: Can I get consensus thoughts on this post? It's a (wrong) answer to the question and redirects to a user-uploaded file download site with lots of ads to install mal/bloat/something-ware.
2
 
@DanielWiddis Good thing I applied a heavy spray of Dramaway(tm) to that post so absolutely nothing dramatic can happen there ;-)
 
@DanielWiddis I'm actually curious what the right choice is there, too. On the surface level, if the download does indeed contain what the question asker wanted, then I suppose it does answer the question. It doesn't look like the website is operated by the user, and while the ads are annoying they (probably?) don't promote a service made by the author.
 
9:11 PM
@DanielWiddis Interestingly, it does not appear to be itself malware
 
Upon talking this out (rubber ducky ahoy) I'm leaning FP.
 
@Spevacus right, but does the user possibly get "credits" for downloads or something? I don't know how these sites work.
 
That's a fair point to raise.
 
I'm obviously not about to run it, but 62 antivirus systems at least all agree it's not malware.
 
9:12 PM
!!/report blender.stackexchange.com/questions/135366/… "Posted on Sept. 15"
 
[ SmokeDetector | MS ] Manually reported answer (94): Does Blender offer network rendering on a LAN by Rabit on blender.SE
 
@RyanM it's certainly not the "latest" Java version :)
 
tpu- feedback received
 
@SmokeDetector Video mentions "Rabit from StateOfTheArt here" in the first few seconds. I think it's fair to say there's undisclosed affiliation.
 
9:13 PM
:55566991 That's fundamentally not true. If you weren't trusted, you would never have been given access. Until such time as you persisted in saying that you would do something we very clearly stated we didn't want to happen, I trusted that you would bring such ideas forward and accept the decisions made. Your statements earlier today make me not trust that right now.
 
@RyanM well I left tp, I'm open to convincing otherwise. But if split-feedback is a system design, I'll leave it split.
 
@DanielWiddis Yeah, my feedback would be a very hesitant fp on the basis of benefit of the doubt. But I'd be open to convincing otherwise, too.
 
9:16 PM
!!/watch billnye\d++@yahoo\.com
 
> Would not be caught as a post, title or username.
 
@user12986714 You don't have blacklist manager privileges, but I've created PR#4851 for you.
CI on 1f058bc succeeded. Message contains 'autopull', pulling...
 
@RyanM my thought is there's probably some incremental benefit similar to "youtube views" where someone is motivated to redirect traffic to their download. They may get a cut of the malware browser ad views, for example.
 
PR#4851 ("user12986714: Watch billnye\d++@yahoo\.com") opened by SmokeDetector
 
9:16 PM
[ SmokeDetector ] SmokeDetector started at rev 1f058bc (ArtOfCode: Formatting --autopull) (running on teward/Osiris, Python 3.6.9)
 
Restart: API quota is 7158.
 
@metasmoke @user12986714 what makes you say this is tp?
 
9:18 PM
@RyanM my-husband-email style trolling I think
Let me dig some other posts out
SELECT id, is_tp, is_fp, is_naa FROM p_posts
WHERE body LIKE '%my husband%'
AND body LIKE '%email%'
@RyanM Run ^ in the SQL explorer should give some example
 
user435118
@SmokeDetector @ArtOfCode I'm pretty sure that won't change anything
 
@Daniil At this point, I'm not sure. At this point, you will need to convince the other admins, in addition to me.
 
user435118
@Makyen Well, how can I do so or even communicate with the admins?
 
This chatroom still works... or there's the email address
 
9:24 PM
@Daniil Saying you definitely won't do it in any uncertain terms without explicit permission from the admins would probably be a great start
 
@metasmoke misclick
 
Was about to say! ;)
 
@user12986714 hmmmm, interesting. It'd be a new tactic for them if so, but I see your point, it'd be kind of weird to include the "hacker's email" otherwise. On the fence here.
 
9:30 PM
sdc approve 4850
 
user435118
@Rubiksmoose I never actually said I'd do it though...
 
@Daniil May I suggest that you give us the tests you want to run, so that we can verify their accuracy on historical data, and then work from there? Instead of sending automated feedback? I think part of the problem here is how vague you're being with what you want to be doing.
 
9:31 PM
@Daniil You implied it heavily. You also still haven't said you won't do it.
 
Automated feedbacks kind of just don't make sense for our project, because the software we write kinda does it's own tp/fp evaluation before sending things to chat in the first place. If there's something automated that would help with that, it should live in the software.
 
@Daniil Yes, you did:
1 hour ago, by Daniil
@Makyen Will if there's 99.99+% accuracy I won't review by hand.
That, combined with all the other statements, wore down my trust that you won't do it behind our backs. It particularly didn't help that you mentioned, more than once, that you expected us not to be able to know if your bot was giving automated feedback, or not.
 
@DanielWiddis Could I get an admins thoughts on an old undeleted fp auto-review job? It would be an automated MS job that finds posts that a) have 1 feedback which is fp b) still be undeleted on-site and c) be older than 30 days?
 
tp feedback received on [MS] Did FASCISTS use disposable plates?
 
fp feedback received on [MS] Did FASCISTS use disposable plates?
tpu- feedback received on [MS] Did FASCISTS use disposable plates?
 
@NobodyNada Unless I misunderstood, it seems like my criteria are slightly stricter?
 
@thesecretmaster They are the same I think
 
> Automatically review all posts older than 1 month, have no true positive feedback and is not deleted
 
9:38 PM
@thesecretmaster I think I added "unedited" in mine. (Also I'm not an admin, just a stats nerd.)
 
@thesecretmaster nevermind, your criteria includes "has 1 fp" while @user's doesn't
 
Ah ok, seems like that's been veto'd then.
 
user435118
@thesecretmaster I don't even know myself fully. I just suggested something and everyone's reacting to it like I said I'm gonna destroy everything.
 
@thesecretmaster I don't think so? It's under active discussion at that link.
 
@thesecretmaster Could be rolled into @DanielWiddis' suggestion, I think, of allowing old posts to complete review with 1 feedback
 
9:39 PM
@metasmoke @Rubiksmoose that's a known troll
 
user435118
@ArtOfCode Email is terrible.
 
well yes
no disagreement there
but it's an option
 
@thesecretmaster Thank you for adding the "no feedback" checkbox on search. Would it be difficult to also have a "1 feedback" checkbox?
 
@Daniil Then there is generally a problem in communication I think, just like how I get blocked from MS
 
user435118
9:40 PM
Who are you responding to? @ArtOfCode
 
@Daniil Maybe step away and figure out what you want to do, and then come back with a fully-formed proposal?
 
@RyanM Yeah. But does that mean it gets tp? Flags yes, but I can't expect the system to actually learn anything from seemingly normal discourse right?
Do we tp those?
 
user435118
@user12986714 Not a problem with the email, just email in general.
 
@Makyen Probably, but I won't be able to do it today so could you make an issue?
 
@Daniil those three messages? to you
 
9:41 PM
@Rubiksmoose I really don't think that's a remotely serious question. They have another post that's not a bad one, but this definitely is trolling.
 
@Daniil I don't think we are talking about the same thing
 
user435118
@thesecretmaster I already come up with one: everyone stops treating me like I'm gonna destroy everything.
 
> I just suggested something
 
@Daniil Chat is a horrible format to try to put together a coherent proposal with statistical validation. Perhaps as @thesecretmaster suggested, making a clear post with your proposal would go a long way to us understanding. Your chat messages are not making your intentions clear to us.
 
@RyanM It absolutely isn't. Not even under debate. But does it help the system to go tp? I'm struggling with that logic.
 
9:42 PM
@thesecretmaster Sure. If it's within the scope of (mostly) copying what you did for 0 feedbacks, I could try adding it, if that would be easier.
 
@Daniil Ah, I meant a proposal for how you'd automate feedback, not for interpersonal stuff
 
@Rubiksmoose It helps future human reviewers, for one thing
 
@Rubiksmoose yep, TP it for ADC
 
user435118
@thesecretmaster When I come up with it I'll do so.
 
It helps people who aren't here for this live discussion and might have input.
 
9:42 PM
I'm open to it (you can see from chat that I started with tp already) but I switched
 
@thesecretmaster If this kind of things keeps popping up, we may just want to give core users the ability to do raw sql search on p_posts...
 
@NobodyNada Good point.
@ArtOfCode Can do.
 
user435118
In the meantime feel free to restore the numerous places you've removed me from.
 
@user12986714 the SQL exploder lets you do that already
 
@Makyen I think I considered it at the time and it added a smidge of additional SQL I didn't want to write. It should be very similar, just a count(*) = 1 rather than a is null
 
9:43 PM
@Rubiksmoose It also helps in case we are considering watches of terms like "fascism" or whatever
 
@ArtOfCode Not with the fancy UI
 
@user12986714 huh?
 
I want a list of titles with feedbacks listed for each
 
12 mins ago, by Rubiksmoose
@Daniil You implied it heavily. You also still haven't said you won't do it.
 
@NobodyNada ah yes, that makes sense.
 
9:43 PM
And clickable links
 
@Rubiksmoose Even ignoring the fact that that person is banned on multiple sites and is often evading that ban (not sure if they are here), I think we generally consider trolling tp (and actually that person is why that keyword is blacklisted)
 
And spammer username
 
@user12986714 yeah, you can do that
 
user435118
@NobodyNada I never said I will do it.
 
@ArtOfCode Technically, but it requires effort
 
9:44 PM
It would if you could do it any other way as well...
 
If I want all the features of the search UI, that will be significant effort
Much more than just the main select clause
 
So what are you thinking of?
 
@RyanM I'm open to correction on this, but generally I vote based on the post itself and not on whoever made it or what the circumstances are. Obviously this is one exception, but am I wrong to take that general approach? (I'm fully aware of this troll and its history and have been very active in destroying them for quite a while FWIW)
 
@Daniil you have also never said you wouldn't. Making a clear statement that you won't would be very helpful. It has been suggested already. You do not seem to think that's good advice.
 
@Rubiksmoose Generally, yes, that's a good approach - known trolls are an exception - ADC, asadefa, etc.
 
9:45 PM
@ArtOfCode Generally, have a search page with a 'sql' input box, and let user select post ids; after that, present those posts the same way as search results (which have fancy UI like spammer name, post creation date and existing feedbacks)
 
@user12986714 so what would you put in this theoretical "sql input box"?
 
user435118
@DanielWiddis Fine, I won't.
 
as in, what's the literal string you would enter?
 
@Daniil I think it's been explained a couple times already how what you've been saying reallllllly reads like you intended to do a thing that you were explicitly asked, multiple times, not to do. That thing is exceptionally controversial and pretty much goes against the philosophy of how feedback works.
3
I think that it makes plenty of sense why the admins revoked your access and do not want to restore it. In this project, access ~= trust, and by implying that you'll do something that you've been asked not to do, even after being corrected multiple times, you've lost that trust.
 
@ArtOfCode Wouldn't any trolling be TP? Like if I were to post trolling on SE it'd still be TP despite the fact that I'm not a known troll.
 
9:46 PM
@ArtOfCode SELECT id FROM p_posts WHERE body LIKE '%string1%' AND body LIKE '%string2%' AND feedback_count = 1 kind of things
 
@RyanM prooobably, but trolling from not-known-trolls can be harder to detect
 
@RyanM I've definitely fped non-serious posts because they weren't spam in the past.
Anyways I have tped the troll.
 
TPing known trolls as well as rising trolls helps to create a history of their trolling that's easy to link to for mods as well, I think.
 
@user12986714 so at that point... why not just use the exploder? Get it to spit out basic information and a list of links, and click on the ones you want further details on.
 
user435118
@Makyen You never granted me any of my MS privileges so it's not very relevant.
 
9:48 PM
[ SmokeDetector | MS ] Few unique characters in answer, mostly non-latin answer, mostly punctuation marks in answer (172): Разряд прилагательных ✏️ by ZaCk - Brawl stars on rus.SE
 
Allowing users to enter raw SQL is asking for SQLi trouble.
 
@SmokeDetector Google Translate says "spam___"
Probably someone edited a spam post
 
From my perspective, I've fped good posts from ADC on sites he's not banned on yet. Once he's banned there, anything he posts is tp because it's ban evasion. Am I doing it right?
 
@ArtOfCode I don't think it is for blazer (so if sql search is restricted to core then it is no problem)
> If I want all the features of the search UI, that will be significant effort
 
Blazer queries have read only access, so blazer is OK.
 
9:49 PM
@RyanM tl;dr IIRC he is network banned
 
user435118
@thesecretmaster I clearly said I won't and still no-one is happy or even acknowledging me.
 
@user12986714 it's not, no, but blazer runs differently to the rest of the application - it's not as easy to perform a query in the same way that Blazer does from anywhere else in the application.
 
@Rubiksmoose oh nice, yay, that makes life easier. is it possible to confirm that?
 
@ArtOfCode Well, can we just pipe blazer result to find_all ?
 
@RyanM I can consult the scriptures and check
;-)
 
user435118
9:50 PM
Huh, I thought my MS account was re-enabled...
 
@user12986714 same again - not easy, because you can't execute a Blazer-type query from anywhere else where you could do that
 
user435118
@ArtOfCode cc @Makyen Can you restore my access, please?
 
@Daniil The thing is, that's not how trust works. You lost it when you continued to imply that you might do the thing even after repeatedly being asked not to. At that point, trust was broken and your access was removed. Getting that trust back takes more than just asking for it to be restored. It takes time and effort to repair.
 
@Daniil okay, I didn't want to make a big thing of this, but you're going to keep on about it if nobody says anything. Your access has been revoked, and will stay that way until your behaviour in this chatroom improves. That's all I'm going to say on that.
 
user435118
9:52 PM
Also, I'd like to receive a copy of all the data you hold about me per the privacy statement.
 
@SmokeDetector v Translates to "spam...." Looks like it was edited (and now deleted) but I'm unsure what the original revision was.
 
@Daniil sure, email that request to smokey@ and I'll get to it
 
Where even is the gdpr page?
 
somewhere?
 
9:53 PM
Ha, made it easy, /privacy
 
@ArtOfCode Ah, how could I have forgotten
Thomas should probably be on that page
 
@RyanM Might be trickier to confirm then I remember. Go with previous info for now.
 
I'm not sure if this has been stated and I haven't read the backscroll... but there's a reason Charcoal caps auto flags at four per post (for the most part) - it's an intentional choice. While y'all have great data and y'all know when something is spam... people still want real people looking at it.
10
 
@Rubiksmoose Thanks for checking, let me know if you find out (or update the known-trolls post). I've been trying to keep it updated with where he's banned but it's hard to keep track and relies on me getting feedback on custom flags.
 
@Daniil I thought you were in the UK, which is not in the EU, so does GDPR apply?
 
9:55 PM
@RyanM I feel you and will do.
 
@Makyen unfortunately yes, until the end of the transition period
 
@Catija As I understand it, this isn't about any changes to autoflags, just changes to how feedback on reported posts works. The proposal that was made was to automate feedback on posts, rather than autoflags.
 
@thesecretmaster similar principle, though - still want human eyes
 
Yeah, but less consequential than autoflagging changes
 
aye
 
user435118
9:56 PM
@ArtOfCode You're not telling me what I did wrong.
 
user435118
@ArtOfCode Full email?
 
@Daniil smokey@charcoal-se.org, as per /privacy
 
@thesecretmaster One compromise that I'd support would be to deprioritize old reviews, rather than getting rid of them altogether
 
@Daniil Full email on this site. metasmoke.erwaysoftware.com/privacy
 
@RyanM It's trickier than that unfortunately. I don't think that one is, but there are... other connections.
 
9:58 PM
Ah, got it. I think I understand...at least partly.
 
@NobodyNada I'm hoping to get some time to work on review soon, and I might throw something like that in - random ordering with a small time-weighted component.
 
@Daniil Does my previous message clarify the problem? As I understand it, it's a trust issue, as I mentioned before.
 

« first day (2495 days earlier)      last day (1328 days later) »