 |
|
You are not logged in! F.A.Q Log in Register |
 |
|
 |
 (nobody)...and 82 guests Last 5 registered Oplandisks nothingstar N_loop yipe foxtrotromeo Browse members... |
 |
Members 8025Messages 2613752 Today 1 Topics 127521 |
|
|||||||||
| |||||||||
|
|
looked like straight from the stage ´84... the blonde locks, the stache, the clothes, sweatband. for a moment i stood there and asked myself if i just witnessed a glitch in the timelime. |
||||||||
|
|
|||||||||
| |||||||||
|
|
metallicadude, top fun with him around the board ya remember?! |
||||||||
|
|
|||||||||
| |||||||||
|
|
It's only now dawned on me that it mermaidman and metallicadude might not be the same person. |
||||||||
|
|
|||||||||
| |||||||||
|
|
all i seem to remember is that it ended in blood |
||||||||
|
|
|||||||||
| |||||||||
|
|
indeed, not the same afaict metallicadude is wavephace. He cleared that up on discord. mermaidman seems like tussle toss maybe..or maybe not...someone old. He won't ever say. He's too stubborn. Will carry on that schtick to the grave. |
||||||||
|
|
|||||||||
| |||||||||
|
|
Tussle Toss is josso. Mermaid man is Tussle Toss. Josso is mermaid man. Josso is English. Mermaid man is Turkish. Dupes are so confusing. |
||||||||
|
|
|||||||||
| |||||||||
|
|
not unlike ethnicity. |
||||||||
|
|
|||||||||
| |||||||||
|
|
i already told you guys i'm tussle toss |
||||||||
|
|
|||||||||
| |||||||||
|
|
we need a zilty brainmap. users are displayed as rectangles of different sizes according to number of posts, dupes have the same color/pattern/pic as original account, lines connect circles according to number of responses or posts within the same topic, one map for every year this place existed with accounts. |
||||||||
|
|
|||||||||
| |||||||||
|
|
^^^that |
||||||||
|
|
|||||||||
| |||||||||
|
|
i could theoretically do this. what's a good thing to generate the graphics, based on the assumption that i can mash the data into whatever it requires as input? because the only thing that comes to mind is React D3, and i refuse to engage with React unless someone is writing me checks. |
||||||||
|
|
|||||||||
| |||||||||
|
|
is there imagemagick for... charts, i dunno, doesn't quite cover it. data visualization would, but that's dangerous; the phrase ropes in all sorts of crap we don't want. no, really, just, say i had all the data, in a phat sqlite db, what's the fastest path towards generating all those rectangles? because how would you even do the layout programmatically? it's not impossible but it's certainly not trivial |
||||||||
|
|
|||||||||
| |||||||||
|
|
oh, dupes. that part is hard too |
||||||||
|
|
|||||||||
| |||||||||
|
|
if, instead of staying up all night writing some php scripts to scrape xlt (at a polite rate, akin to what a human on lots of coffee could manage with a keyboard and mouse [don't want to break the poor old board]), i had a full DB dump... zilty clearly stores IP for registrations, because it only allows one username per IP. or, well, it did, when registration worked... the dawn of the IPv6 era utterly fucked this as a gatekeeping measure, and thus the dupe overload middle of the 2010's. i figure, though, if XLT happens to log IPs for individual posts -- entirely possible -- then once you've rolled through on some cheeky internet config to make a dupe, fuck it, hard part's over, just use your normal home IPv4 to post afterwards, and, from that, you probably could just churn out a list of likely dupe connections with two pages of code and ten seconds of tapping your foot |
||||||||
|
|
|||||||||
| |||||||||
|
|
i guess basic infosec be like: you can use your IPv6 proxy 99% of the time, but just once is enough to connect it with anything else you've posted from just once. but then are these the same person or is it just mermaidman climbing into a window, having a wank on an honest non-dupe xlt poster, then using their computer to log in as a dupe and post bullshit also things like, even if you are using a proxy, you could probably use the IPv6 subnet to pool dupes around likely individuals |
||||||||
|
|
|||||||||
| |||||||||
|
|
while we're here -- i suspect new member registrations could be turned on with the flick of a switch, but that switch was installed and set to off because of IPv6 and other fings. i am fine with this. |
||||||||
|
|
|||||||||
| |||||||||
|
|
in general, i've found this board to be a rewarding spiral of information; charming to think what i could do with more access. |
||||||||
|
|
|||||||||
| |||||||||
|
|
hmmm. even without the IP data you could probably bucket time of day all the posts are made and reasonably guess a part of the world, maybe even a particular time zone. didn't the board used to show that with all the posts? with idm names for all the zones? that may still be there. i'm not sure i'd advocate scraping user profiles or anything that requires login; unless i'm sent the mysql dump should draw the line somewhere. but, now i am thinking, should i snag all the profiles... it's a larger context for the board that i am not capturing |
||||||||
|
|
|||||||||
| |||||||||
|
|
i don't think i have the stomach for trying to programmatically identify possible dupes based on a bunch of algorithms that identify similar writing styles, but certainly also a sideband channel open to the public that's actually a big part of why i never got into dupes myself. i tried, but them it was was like -- wait, shit, i'm writing like myself, i'm going to telegraph it and even without computer science i've seen some people telegraph it. should probably be keeping a list |
||||||||
|
|
|||||||||
| |||||||||
|
|
certainly we cant identify the dupes the maps would still be really nice to have imo also i kinda was expecting someone to draw a mock map ^^ |
||||||||
|
|
|||||||||
| |||||||||
|
|
don't think i'm capable of just tossing off something that informationally complex. just look at what happened when i tried to map all the members of the wtfmm board to characters from Escape From New York i could have just used photoshop or something, but, no, it's a shell script using imagemagick to build the image with a command you can never identify dupes with absolute certainty, but you can comb the data and establish something like "there is a 90% chance these two users are related, even the same person." even without the IPs, you could pile up some heuristics: which accounts post at what times? do two users often show up in a cluster, with posting, then the other, but never seemingly at the same time? post length (i'd be obvious there). average word length. how often commas are used but, yeah, without the IP address data, i don't think it'd really be enough to extract the juicy details i know we all want. with the IP data, though... trick alone with ipv6 subnets would probably rope you a few right off the bat, then if XLT also logs IP of a user logging in, or the IP address every XLT post comes from... i think you could probably shake at least a third of of them loose with code, and human eyes, knowledge of this loony crate, could probably bump that up to half, even. i think this is the sort of shit people use hadoop for. something like tableau is probably more my speed. doubt i'll have time for this, but quite amusing myself figuring out how i'd do it if i did |
||||||||
|
|
|||||||||
| |||||||||
|
|
ip´s are long gone with the wind, but youre angle sounds promising. bet there is some programs out there companies use to gather info about their users already not my field of expertise though |
||||||||
|
|
|||||||||
| |||||||||
|
|
mine either, but it is only a degree or two removed from my field of expertise, and, shit, what was that, i know a thing for this... then i'm googling "big data software" and checking wikipedia and i'm sure you can imagine the trash like "25 top big data softwares" to surf through, then, like: tableau. yeah, that sounds like it. i'm not 100% sure, but it feels right. the one-sentence snippit about themselves they've provided to search engines (they get one, and i'm sure some people in a room somewhere argued over it because search engine rankings and millions of dollars) is: "Tableau is visual analytics software for business intelligence. See and understand any data with Tableau." stitching together statistically likely dupes and "business intelligence" are not far apart at all, and it seems like it's good at generating shiny infographs as well. monthly subscription of $50 or something. i'll read up on it more if i get a chance |
||||||||
|
|
|||||||||
| |||||||||
|
|
i started looking around out of curiosity, and already, this company is a mite irritating. i'm pretty much trying to ascertain: what's the workflow? where's the documentation? and i keep getting bullshit like "select your way to learn." no, not live training, instructor training, no training videos, so finally i click "eLearning" and it's like, yes, would you like the $15 or the $5 plan? so i dial back up and i see "dive deeper, see all our white papers" and i click that and it wants me to register. i try the training videos, and, yep -- please register. i am going to try a bit more to find a publicly available manual, or something, for a bit. then sign up. and if i still get off-script donald trump as a value proposition, well, this may not be the enterprise solution zilty is looking for |
||||||||
|
|
|||||||||
| |||||||||
|
|
so far this almost seems like a glossy layer above the old tradition of slaving across hot spreadsheets all day -- a nice tool for manually digging through a pool of data. i wager you can go more serious with scripting or something, but i haven't gotten that far yet. biggest up so far is they've clearly stolen the chart wizard crown |
||||||||
|
|
|||||||||
| |||||||||
|
|
the "dupe connection" problem -- it's really up for debate w/rt how far you could get, especially with just public data. deep, open-ended, quite interesting. especially when it comes to leveraging what inside knowledge you have in the number crunching. mo has a very specific writing style, for instance, due to english not being his first language. i can't simply cough up a list offhand... but, if i sat there with a notepad, i could probably define a reliable signature, and it would be the sort of shit you could never dredge up with code; it requires human cleverness. a pile of esoterica a computer would never figure out -- like, all i need is his capital letter policy, comma policy, average word length, and then always this verb tense malfunction over and over, many posts... but, even that in hand, how do you program it into the computer? all of those metrics are individual problems, some of which are much harder than others -- like, spotting verb tense mistake, there are entire other ~enterprise solutions~ on sale. just think of how spellcheck is reliably stupid in certain ways; extend that to finding grammatical errors sideband: apologies to mo. i have no problem understanding you. there is nothing wrong with having your own style. and it'd just be a hassle to be a schoolmarm about your writing; i only poke you about it if you made a mistake that gives me a particularly good laugh anyways, i'm not sure tableau is what we want for automated dupe analysis, but it does seem like a good way to, like, you know in jaws, they cut the shark open and start seeing what's in there, a license plate, a car tire, like what could i get out of this pile of nfoz and what's the best way to get it. more like determining what will actually yield results before i program anything. i might spring for the free trial |
||||||||
|
|
|||||||||
| |||||||||
|
|
ahaha, jeez, neural networks. i clearly have all the training data i'll ever need... so you train a network that determines "was this post written by mohammed" and it spits out a probability that is mohammed, 0-100%. you feed it mo posts and loads of random other xlt posts, it guesses yes or no, over a bajillion rounds it has some unfathomable salad of hidden variables. you then test it by running it on all of mo's post (should clear all) and then all of someone else's posts (should clear few to none). if it seems in order, try a few other people. if it's still nailing it, then you turn it loose on every xlt post and see what comes back so there is an argument for nibbling at it rather than trying to blow up the safe |
||||||||
|
|
|||||||||
| |||||||||
|
|
i now have all sorts of silly ideas. like: train NN simulations of all active posters, create a clone of xlt, and it's just a locked-down board with a bunch of simulations of us all replying to each other's simulations. scrape new posts and update models daily dupe thing is a nice thot, a charmingly stupid pursuit, but bottom line, i feel like i have to do something silly with all this |
||||||||
|
|
|||||||||
| |||||||||
|
|
anyways, i'm not sure tableau is what we want for automated dupe analysis, but it does seem like a good way to, like, you know in jaws, they cut the shark open and start seeing what's in there, a license plate, a car tire, like what could i get out of this pile of nfoz and what's the best way to get it. in the end, the shark was a dupe. |
||||||||
|
|
|||||||||
|
Messageboard index
|
|||||||||
