#metabrainz

first number is the number of occurences

this list is seriously WTF.

line 2, line 46

a clearer picture is starting to emerge!

I'm a little surprised that the page with the highest number of hits is so "small" (8000/153978184)

zas: what if you redo it by removing the last component? (so that we get things like /artist, /recording, /ws/js/edit)

I guess some artists take longer to render than others though

Yes, we have a very wide variety of requests, actual number is 8114 over 1017893 (the rest being mostly ws or errors)

ah, that sounds more correct

why do we have so many http2 requests?!?

(and http1.0 requests! are those bots?)

because http2 is much more common now (most browsers support it)

1.0 is very likely scripts

so many spammers…

yvanzo: yes

ok let me few minutes to list only GET requests with prefix as asked by alastairp

yvanzo: ding! spammers.

we've got a ton of spammer traffic hitting us.

zas: can you please run a query to check ws vs web traffic?

used to be less than 5% for web. I think that figure has gone up.

next query: top /user/[name] pages. there is no reason that any user should get more than 10-15 hits in 24 hours.

I think a few tweaks according to what we talked about during the meeting are going to have a massive impact.

e.g. looking at a user page? robots.txt and you must be logged in.

ruaok: yeah, user/bruceluciani4639

skeeeetchy

given that this is a spam problem, not an inefficiency of our code, we need to act a little harder on spam.

Which percentage of spammers among 355 account/edit?

yeah, good questions.

these are spammers editing their profiles.

top 20 GET, stripping anything but first element after initial /

[11:38] <ruaok> next query: top /user/[name] pages. there is no reason that any user should get more than 10-15 hits in 24 hours. +1

top 20, users

221 GET /user/expensiverhythm this also

most to be deleted imho

should we start deleting spammers or wait?

how about keeping a list of these "very often GET'd user/* names and if it hits a certain number message people woh can deleteaccounts right away

each profile has to be checked, but most are prolly editors with no edit, or recently created

and keep doing that

sothat peopel can check them and delete them accordingly

(I kinda don't wnt being able to watching my profile account you have to bel ogged in, i mena i link that all over the place as a n advertisment *for* mb)

yep.

zas: can you please prepare to run these queries daily, with a report that links back to the user profile?

i cna go trought them if you guys want?

erh these zas listed today i mena :)

are there things like akismet for user profiles? does it make sense to perform some kind of validation on profile edits as they're made?

CatQuest: thanks, but lets leave this to reosarevok, Freso

alright!

alastairp: we need one, for sure.

ruaok: ok, i'll look in an efficient way to do it

thx

I guess it's clear that these are being made manually? since we have captcha

oh no pita about not adding url to my profile tohguh.. i'll be ok with it if "validated" users lie kme and others who literally edit alotds coudl add links and stuff

when I say "validated" I mean, autoeditors, all theditors that make loads of editors like hibiscus etc

alastairp: or automated captcha solving.

zas: https://www.irccloud.com/pastebin/jmS4sX3F/ contains duplicates.

not the same protocol

woo, it feels great to finally have some insight.

zas: for this query we don't care about protocol, please collapse them.

ok

collapsed, top 50

great, thanks. Freso, reosarevok. Can you please help out?

we need to implement the suggestion from last nights meeting ASfuckingAP.

and we need a report that lists all users who have a bio/link, but zero edits.

I remember when.. ian showed me numbers /graps and said that they were spammers

and I explaimed we shoudl delete thme

and he didn't (and noone esle did either) care sicne they jsut sat there and did nothing just inflating user count

thinking out loud here... if captchas reduce the speed at which spammers can make changes, what about putting them on all (or many) forms? profile edit, collection create/edit

yup, now that policy is haunting us.

ruaok: they could have critiquebrainz works thoguh

alastairp: how aobut NO

google has the invisible captcha now. let them trigger it when they think the user is spammy

that is a much better idea

++ for that

(can they report to us how many requests triggered it?)

I doubt it.

making it harder to edit the profile is a good starting point.

removing a pile of users would also help a lot

it doesn't make sense to add it if we can't see the impact that it has on changes

this is one case where being right all those years ago does not please me :/

yeah, agreed.

lots of people have shouted me down on spam,but I am going to take a much harder stance now.

at least all of this is finally starting to make sense. NONE of the entity pages or otherwise MB centric stuff is showing up in the top.

just! don't make it harder for legitimate users editing things :x

CatQuest: I don't plan to. but I don't plan to let people shout me down from taking action.

am I shouting?

i'm agreeing with you! infact I say "welcome after!"

no.

alright then :)