Charcoal HQ

@ArtOfCode If there's something people not in it should be taking away from it, could someone summarise the pertinent points of what happened & what we've learned together?

[ SmokeDetector | MS ] Offensive answer detected, blacklisted user: Why were the ways of the Force considered a "religion" in the Galaxy Far Far Away during ANH by non-Force-users? by Radon the guy on scifi.SE (@Mithrandir)

tpu- by Catija

From reading the tail end of it, I gather there's something to be done about interactions with sites, Smokey being active there, and those sites having capacity to opt out of specific features.

Also do I have the right idea of what needs to go into those manuals?

Give me some time and the right info and I'll help you write a lot of it.

Hmm does MS no longer flag for me per this 20 mile transcript? Or do I just need to reauth again?

metasmoke.erwaysoftware.com/flagging/logs?filter=manual

DJMcMayhem

!!/report scifi.stackexchange.com/a/182632/50357

12:09 AM

[ SmokeDetector | MS ] Manually reported answer: What is the concept of androids in Doctor Who? by Radon the guy on scifi.SE (@Mithrandir)

@doppelgreener I'm writing a thing up about the conversation and stuff, I'll send it when it's done (later today/tomorrow)

tpu- by K.Dᴀᴠɪs

Yup, just an auth token issue (again)

@thesecretmaster thank you very much. i appreciate it -- i wasn't present but it seems like something important for those of us who weren't to be across.

@SmokeDetector ahh so we completely disabled AF? That was a high score

@SmokeDetector autoflagged

12:11 AM

@K.Dᴀᴠɪs That post was not automatically flagged by metasmoke.

@K.Dᴀᴠɪs I was under the impression that we don't autoflag for R/A?

You're correct. Good eye

I am still dazed from reading the transcript :P

API quota rolled over with 7648 requests remaining. Current quota: 19999.

math: 61
stackoverflow: 37
physics: 16
superuser: 13
askubuntu: 7
drupal: 7
history: 5
security: 5
codegolf: 5
scifi: 4
mathematica: 4
electronics: 3
apple: 3
bitcoin: 3
english: 3
dba: 3
datascience: 2
webapps: 2
rpg: 2
hermeneutics: 2
wordpress: 2
ell: 2
interpersonal: 2
unix: 2
patents: 2
salesforce: 2
magento: 2
serverfault: 2
puzzling: 2
stats: 2
android: 2
academia: 2
es.stackoverflow: 2
workplace: 1
ru.stackoverflow: 1
music: 1
monero: 1
cooking: 1
sqa: 1
gaming: 1
aviation: 1
chemistry: 1

[ SmokeDetector | MS ] Bad keyword with email in answer, email in answer: How to become a vampire and werewolf hybrid? by vampiredefilers on gaming.SE

tpu- by K.Dᴀᴠɪs

[ SmokeDetector | MS ] Repeating characters in answer: What’s the correct way to refer to “the Doctor”? by Ftdftkdijjstiofuoldisigurausis on scifi.SE (@Mithrandir)

tpu- by K.Dᴀᴠɪs

@SmokeDetector oh boy, arqade gets vampire spam

1:32 AM

Spam over time, last two yearsish

January this year was the first one over 4500

Random thought: Maybe the paradigm in Charcoal should shift from "destroy all spam" to "help sites stay clean." That attitude might help with conflicts between Charcoal and mods/the rest of SE.

J F

1:58 AM

Yikes that’s a long transcript

Funny how there’s almost (?) no Smokey reports during the argument

(or have I just trained myself to become blind to them?)

If you don't want to read the transcript, I'm attempting to summarize it.

@thesecretmaster "Clean" seems somewhat ... non-specific.

I mean "spam/garbage free". I was thinking that the helpful bit would be changing from "destroy spam" to "help communities destroy spam"

I'm just tossing out an idea, I really don't know if that would be good/useful at all.

John Dvorak

2:20 AM

I don't see much difference there

It's interesting, but needs fleshed out - deleting spam is a pretty simple thing. I don't know what value communities can bring to the process

The answer to "what do communities do with spam that we can't" is what you'd need for that plan.

... I think that's sort of the point. There's a difference between "We'll do it for you (just sit back and forget it's a thing)" and "We'll help you do it (let us know if you need any help)"

It may not be an effective difference, so much as an optical one.

It's all about framing.

I guess what I'm trying to say is that I think the attitude should be less "We're going to help you whether you like it or not and we're going to help you our way" and more "How can we best work with you to keep your site spam-free"

It's the difference between telling your spouse "Your face makes time stand still" and "Your face could stop a clock".

I'm thinking of it more as an attitude change than an actual process change.

Basically what @Catija said -- it's an optical change, but that change might foster better, more cooperative relationships between Charcoal and mods/communities.

2:30 AM

Want to write up a paragraph in a gist or something? My main concern is that nothing would change without some kind of hard change. But I could be wrong.

I think I'll put it in my gist summarizing the above conversation. I'm working through that slowly now.

Awesome

@Undo I think you're right, it would be hard to do without either a hard change (and I don't know what that could be) or simply Charcoal members voluntarily choosing to approach things with that attitude, and some kind of official statement about it for people to point at.

My concern is this: What happens the first time a site asks us to catch NAAs, or tweak a reason in a way that doesn't make sense in some way? Do we still hold our ground on that, or do we start compromising on technical matters?

@thesecretmaster I like the sentiment, but "help sites stay clean" is a lot more ambitious than removing spam. "Clean" can be related to quality and quality is domain specific and even in that it's an issue that sites have struggled with internally for a long time. "Garbage free" has the same problem. On SO there are at least monthly arguments about how we let garbage on the site

2:34 AM

Yeah, clean or garbage free isn't really what I meant. I'm not trying to suggest a significant process change like that.

@Undo NAAs are out of scope. We don't have the domain knowledge to know what counts as NAA for them. We're here for spam. Period.

I've argued against NAAs and non-spam catches in the past because it becomes very slippery very quickly. Some of those are side effects of what we do, but seeking out non-spam should not be our goal

Ah

@Catija "But you said you were here to help"

One of my recent complaints about FIRE is that if you use it to mark something as Rude, it increments the Spam number, not the Rude number.

@Undo That's why the wording is specific and spam-centric.

I dunno - just seems like we need to pick something we can do for sites on request. Otherwise it feels empty.

2:36 AM

That's why I balked at "clean".

@Catija that's because spam and rude send the same feedback

@ArtOfCode But when someone else uses fire, they may not bother actually paying attention and they may go with spam instead of rude.

... because it's incremented.

I've close voted stuff based on looking at what other people have close voted as... it's natural to find consensus...

@Undo If they ask you to flag NAAs, you can say (nicely) 'sorry, that isn't in our scope,' and maybe point them at other NAA detection tools. For reason tweaks, you can say "Well, why do you think this tweak could be helpful?" and try to work with them to agree on an acceptable tweak.

@Catija that shouldn't matter. They do the same thing - only difference is that spam actually casts a flag. Which also shouldn't matter, because spam flags should be accepted on R/A posts

@ArtOfCode But it doesn't. It's classified differently in a user's profile. Behind the scenes may be similar/identical, but the way SE presents it to a user they are two very different things

2:38 AM

@ArtOfCode But the internal end of things does mark them differently.

@Andy I'm talking about feedback to metasmoke - in FIRE, the spam and rude buttons both send tpu to metasmoke

We need to mark things correctly. Period. Spam flagging R/A because it's convenient isn't a solution.

@ArtOfCode But the spam button flags as spam... the Rude button does not flag as anything.

^

We're not really going to have this argument, are we? This is something that we've been agreed on internally for ages - spam and abusive flags should not be getting treated differently.

We need to stop focusing on OUR tools and remember that the most important thing is what happens here, on SE.

2:40 AM

@thesecretmaster So the same thing we do now, with less teeth and more invitation. Could be a good principle.

Feb 28 '17 at 21:31, by Pops

It's... hard... to argue in favor of writing an algorithm and making a bot spend time calculating which of the two flag types to use when they are, in actuality, the exact same flag. It's not even Coke vs. Pepsi, it's Pepsi vs. a second bottle of Pepsi that's using a different but still valid Pepsi logo.

[ SmokeDetector | MS ] Few unique characters in body, no whitespace in body, repeating characters in body: Javascript string used in function randomly getting messed up by nino on stackoverflow.com

A: Spam flags and Rude/Abusive flags - Is the difference practical?

8

No, there is no practical difference between the two. The only reason for keeping them separate is the different signals that each sends to us. To expand on your first point, yes, spam flags can trigger a review audit. We explicitly don't include abusive flags because we don't think we should be...

@Undo Yeah. Again, I'm thinking of it more as an attitude change than a practical change. The problem I'm trying to think of a way to address is the way some mods/sites have negative impressions of Charcoal. It seems adversarial when it should be cooperative.

It'd be nice for FIRE to offer that option, but it's not high priority.

2:41 AM

@Undo then we get SOBotics to set up natty

right tool for the job

@quartata There we go

tp- by thesecretmaster

If charcoal thinks of itself as "helping sites destroy spam" instead of "spam crusaders," I think that it would foster a more cooperative attitude.

@Undo i would add that if we really commit to Inferno it would make charcoal and sobotics a bit more conjoined

aside from the fact that would make starting natty on all sites trivial as opposed to impossible like it is now

[ SmokeDetector | MS ] Bad keyword in answer: Free screen recorder with no watermark/limit by Simonhong on softwarerecs.SE

2:51 AM

I have my reservations about Inferno. I think it's a good project to have, but I don't think it's going to work for the Smokey use case.

why?

you should probably be telling me these things now rather than later

fp- by thesecretmaster

because I want it to be as flexible as possible and if it doesn't work for us it probably won't work for most

the SOBotics bots actually have way more complicated requirements, especially when scaling out to all sites

@ArtOfCode if this is about autoflagging I actually have plans for that -- we're not the only bot that autoflags, after all

@quartata no, it's more general than that.

metasmoke is a specialised tool that's very well adapted for our use case. It's a thing we can add literally whatever we want to - reports, data, stats, tools, whatever. It does what we want, in the way we want it, and it's adapted specifically for Smokey because that's how it's going to work best for us.

I have reservations that Inferno is going to be able to replicate that. Simply by its nature of being a thing that's not adapted for a specific bot, we're going to lose a bit of finesse. When it comes to wanting to add new things, we're also going to lose out because we'll have to think of how to adapt them for other bots as well - and that's significant work.

my intent is that Halflife will fill the roles on these things that aren't useful for other bots -- currently that's exclusively domain tracking

pretty much just creating a web dashboard for some of the logic that's already there

3:02 AM

This whole project thrives on flexibility, the ability to do what we want, how we want it, unconstrained by anything but ourselves. The best way to do that... is to write our tools specifically for it. Anything else is going to lose us something.

[ SmokeDetector | MS ] Username similar to website in answer: Import my SSH key as GPG sub-key to use for SSH authentication by cuihao on unix.SE

And, of course, back to the old "if it ain't broke". metasmoke is very much not broke.

@SmokeDetector I thought that only caught users with 1 rep?

@Catija up to 50

8 hours ago, by Undo

@terdon Some context on how that reason works: It's only active on 1-rep users (like most of what we do), so that guy will never be caught again anyway

... so... yeah.

3:05 AM

that was... incorrect

So maybe that's something else we should be codifying in our documentation somewhere...

Quick reference access to triggers so that we don't have to go by memory.

fp- by Rob

Possibly, but that would be a nightmare to keep up to date

Do they really change so often?

Too bad there's no way to reference the code in the documentation... that'd be pretty cool.

@Catija "reference" as in how? You can always link to specific lines

3:10 AM

In theory that would be expressed by tests

@ArtOfCode "highlight" the number and refer to it in the documentation... like referencing a cell on one page of a spreadsheet from another.

I do think we should start teaching people more about the code internals

Even if they're not contributors

@Catija yeah, that's... less easy

@Catija actually

this could be doable

@quartata Hey, fancy writing a training course? ;)

J F

3:11 AM

@thesecretmaster no I read it.

we could make something that generates docs from the reasons list

I'm writing basic training, but an Internals 101 thing would be pretty awesome

J F

A summary would still be useful for the future.

@ArtOfCode all right

I guess it would go in our channel?

@quartata I was thinking on the website along with the basic stuff, but it could go in both

3:12 AM

@quartata No. Users don't care about the code behind a thing. They care that it works and does what they want. They care about how to do things in the system but not in the code.

Heeey, I know things about teaching CS stuff. Maybe I can help write this.

@Andy the average end user

this is different

@thesecretmaster that would also be awesome

many of our core contributors contribute in ways other than actual code

that doesn't mean they shouldn't know a little bit about it

that's the audience

@quartata Is it? Who is your audience in this case? The mod that comes in with a question or request? A correct answer is not "Go look at line X on GitHub"

3:13 AM

@Andy Agree that we don't need to teach literally everyone about the code, but having the guide there for people who want to know about it would be good.

Does python have a tool like YARD/rdoc?

@Andy look no further than right now for your audience. if Art hadn't been around, that question could have gone unanswered

@thesecretmaster yes

That's two different types of things - both of which are good. But, if we're going to document stuff to point to as reference to answer questions, that's not code internals

teaching people where to look when they have these questions is at minimum what I'm thinking

@quartata Does smokey use it?

3:14 AM

and of course teaching more for people who are actually going to contribute is good too

@Andy But we're talking about putting something together so that when mods or other users have questions, we can answer them accurately. Undo told terdon that the limit was 1 rep... but it's 50... that may not do harm but it's not correct, either.

we should target both

@thesecretmaster hell no

That would be the way to do this, IMO. Otherwise the docs will get out of date sooo fast.

@quartata @Andy you're saying the same thing. A thing for teaching people who want to know, or who want to be able to find things out for themselves (like the reason max rep). Not for the mod who comes in with a request.

@Catija Ah, so the goal is for us to know where to go to get the answer. It's not so that we can drop a link to github for a non-technical person to parse?

3:16 AM

@quartata Cleanly written, it can achieve the same. Yes. If it's going to help an outsider, it needs to be free of jargon...

@Andy I think that would be useful... yes. That might not be the entire goal but it's a big part (for me).

Cool. I'm on board with that

Take, uh... @Mithrandir as an example. Not a developer, but very much involved in this thing. It'd be useful for people like that to know where to find things in the code - or, if they don't know, to have a place that will tell them where to look. That's Internals 101. Internals 102 is the more detailed thing for developers who are looking at working with the codebase.

Especially if it seems to change so often... it's good to have somewhere it's up-to-date since memorizing it would be a waste.

@ArtOfCode Mith is the perfect example

Mith is officially the spirit animal of this effort

4

Update on the summary thing: I just finished going through the whole transcript and taking notes. I'll have reasonably neat daft in ~24 hours.

3:23 AM

You deserve a cookie @thesecretmaster

but not just any cookie

A good cookie

CM auth cookie for SE?

HAHAHAHAHA... I doubt it.

3:24 AM

maybe to good of a cookie

awwwhhh yeah

Do you really want to have to accept the mod agreements on 170 sites?

... twice each...

@quartata Mmmm! Chocolate with a bit of cookie underneath.

hahaha that's a fair description

Berger Cookies

Berger Cookies are a kind of cookie made and distributed by DeBaufre Bakeries. They are topped with a thick layer of chocolate fudge that derives from a German recipe, and are a cultural icon of Baltimore. == History == The Berger Cookie recipe was brought to America from Germany by George and Henry Berger in 1835. Henry owned a bakery in East Baltimore that was later run by his son Henry. While the younger Henry took over his father's bakery, his two brothers, George and Otto, opened their own bakeries. Around 1900 Otto died, then George and Henry combined the bakeries to create 'Bergers'. As...

@Catija I really just want to see the SO flag queue with my own eyes :)

3:26 AM

[ SmokeDetector | MS ] Manually reported answer: How to save data from text box to mysql database using asp.net (vb.net) by Luke Pedero on stackoverflow.com

tpu- by thesecretmaster

Well... that doesn't require a CM diamond, silly... SO diamonds are much easier to get... there's 25 of them... there's only 12 CMs.

10-second brainstorming, y'all - answer this question: "What is Charcoal?"

The stuff I put in my barbecue?

capital-C Charcoal ;)

@Catija "Undo, wanna donate a cookie?"

3:27 AM

@ArtOfCode we detect and analyze spam across SE

@ArtOfCode Helping 170+ communities network wide detect, track and eliminate spam.

A group of Stack Exchange users working together to find, analyze, and eliminate spam from the Network.

@ArtOfCode Maybe turn that into "What is Smokey and surrounding systems." We're just the group that maintains that and a few other things.

you know

it's recently occurred to me

I have no idea what the original Charcoal did

other than that it was in PHP...

comments, I believe

3:36 AM

@quartata Uh oh. Was it really?

github.com/Charcoal-SE/Charcoal

@Undo yeah, that was deliberate. Training docs are going to start with what the organisation is, then jump into the systems.

I think I'm going try to host an instance just to see it

Makyen

For the message from SD of "fp feedback on autoflagged post:" could we please get a link to the actual SE post. Personally, when notified of the issue I'd find it much more important to get to the actual SE post rather than the MS page about the SE post. After all, the entire point of that message is to get the autoflaggers to personally evaluate the SE post and chose to retract their flag, if appropriate. It can also link to the MS report, but the SE post should be the primary link, IMO.

@Makyen Makes sense; wanna throw an issue on the MS repo so I don't forget to do it?

Makyen

3:42 AM

@ArtOfCode Sure

3:53 AM

CI on de99f5c succeeded.

4:03 AM

[ SmokeDetector | MS ] Few unique characters in body: How can I replace part of a string with a pattern by yibs on stackoverflow.com

[ SmokeDetector | MS ] Mostly dots in answer: What does the notation $(\frac12)^3_n$ mean?? by A guest of math on math.SE

tp- by Makyen on How can I replace part of a string with a pattern [MS]

tp- by Makyen

Makyen

@JakeSymons I haven't tested it, but the script Retract flags from the “Flagged Posts for [User]” page appears to have been made in direct response to someone else desiring to retract flags on deleted posts.

4:19 AM

Grr... why isn't the query string preserved when you click a link in rails

@thesecretmaster Is that expected behavior in any web language?

Doesn't seem intuitive to me

I don't know if it's expected, but I didn't expect it.

I kinda thought browsers automagically did that, but meh.

If it's a link_to, I believe there's an easy to to put the current params in the query string

@quartata Obsolete comments. Which is why I push the distinction between 'Charcoal' and 'Smokey projects'

For example, we theoretically also maintain Blaze.

Smokey takes up most of the time, energy, and discussion, but at the core Charcoal is just a bunch of people at the intersection of SE literacy and technical chops.

@Undo I googled and couldn't find it. I was hopeful that **params would work, but it gives an UnfilteredParams error.

Is this a metasmoke thing?

I think I do it there, somewhere.

If you're not concerned about XSS, stackoverflow.com/a/9963335/1849664

Or stackoverflow.com/a/23578450/1849664

4:32 AM

It is metasmoke. I wanted to get a start on the site dash idea that's been floating around.

Oh, awesome

And you need to keep params around to do filtering, sorting, searching, etc

Yep, exactly

I'd definitely look at stackoverflow.com/a/23578450/1849664 and just make sure you validate the heck out of params in the controller.

Actually... I think it's an issue with name_path(some_param: value)

And form submission with GET

Why do you need a named path?

A link in the ?order=foo form should work to load the current page with that param

4:35 AM

Oh wait... links... ugh.

Yeah, I made a stupid.

Oh no I didn't

I'll just keep doing it the annoying way for a bit, and you can correct me once I've got the PR created.

Next question: What should and shouldn't be reported to Smokey?

Spam is an obvious yes

NAA is an obvious no

Questions on R/A and vandalism

R/A yes, vandalism no

The "report red-flaggable things" principle, combined with the "would we block this at the SE level" principle for feedback, has served us pretty well.

Would graphs be overkill/too expensive on the site dashboards?

Depends on the graphs, but it'd probably be fine

We can always scale back if necessary, but these views probably won't get much traffic anyway.

OK. The graphs in /graphs are global, so I'll look around for some local graphs I can use.

4:42 AM

Maybe the post volume + tp/fp graph, a TTD graph, anything else?

@thesecretmaster You could let the backends for the /graphs ones take am optional param to specify a site

We can also cache graphs

I wrote a cached_query helper somewhere

*an. I can't type.

@ArtOfCode You already do, so I think that making it per-site would screw that over.

Oh no it wouldn't

Nevermind

OK. I see how to do this now

@ArtOfCode Want response to the other thing email? Still relevant?

Some encouragement after today: We're in entirely uncharted territory. I think we're doing fairly well, all things considered. If our biggest problem is social relations with single or double digit numbers of people... that's a pretty standard problem for a bunch of programmers. This is a course correction, not a "we're doing everything wrong, turn the ship around" thing.

4:47 AM

@Undo Probably worth it, if only to work out a response for any future requests

eh, okay

@Undo Vandalism is a side effect of what we see though.

on the other hand... eh

I'll leave it up to you :P

Actually, I can take part of it here. After musing this today, I realized a big part of the cost difference between disabling a site and disabling or tweaking a reason.

Disabling a site is rare, can happen in one place, and is pretty easy to comment.

If we disable or tweak a reason, we've now made a promise to freeze at least some part of the code. Communicating that - especially if it's a tweak, not a disable - would be difficult on a team this size.

So I'd be okay with a "we reviewed this, decided to turn it off but we can turn it back on if we think it's best"

A "yes sir, we disabled that reason" is far harder to maintain. Now you're dealing with constraints beyond optimizing for accuracy and volume, and everyone has to know about and remember those

I get what you're trying to say, but... code comments?

4:54 AM

[ SmokeDetector | MS ] Linked punctuation in answer, non-Latin link in answer: Problems with file path by davidb on stackoverflow.com

@ArtOfCode There are other ways this happens. Are we also obligated to ask or disable future reasons along those lines? Suppose we get our ML act together at some point and get a "Site affiliation not disclosed" reason. Now we have to remember which sites didn't like related reasons and discuss whether to make them opt-in, opt-out, or something in between.

fp- by tripleee

fp feedback on autoflagged post: Problems with file path (@iBug)

Or suppose we started loading user profiles, and got a "website in answer similar to profile description" reason. It would be reasonable to say that BI wouldn't want that reason... and now we have to remember that.

I'm saying BI here assuming that a meta post gets made, and that community consensus is "no".

Hmmm... I think I'm staying on the fence at the moment. I totally take the point, but disabling the whole site when Smokey is actually much more fine-grained just feels somehow wrong

I think it's possible to solve the issues with it, but the debate is probably more whether it's worth it