Monopoly Man – the same image uploaded to Imgur 3.2M times? | Hacker News Reader

Monopoly Man – the same image uploaded to Imgur 3.2M times?

imgur.kosiru.com

321 pointskolbe69 comments

Interesting, but the analysis is a bit short. This was posted to Reddit five months ago with more discussion:

https://www.reddit.com/r/self/comments/2uzas7/i_found_out_th...

They found (as I did, with a simple Google reverse image search) that the image was used to illustrate a Forbes article in 2011:

http://www.forbes.com/sites/venessamiemis/2011/04/04/the-ban...

https://twitter.com/EricaGlasier/status/563770191440404480

Doesn't actually answer the question, but the source of the image is I think super relevant.

I could sort of imagine some automated process making Imgur thumbnails of images from links shared on some social media site, and going awkwardly self-recursive and getting to 3.2M.

Just dug up these links too. This comment on the Reddit thread may explain why https://www.reddit.com/r/self/comments/2uzas7/i_found_out_th...

Interesting speculation:

"Sorry to be boring, but Webdriver Torso already gave us the answer: it's a test image used in integration tests of Imgur itself. Silly image."

https://www.reddit.com/r/self/comments/2uzas7/i_found_out_th...

why would they have test data in production?

Why would they not be testing production?

I guess it makes sense for their case, since who cares an image was uploaded by integration test or it was "real user" generated.

Seems like the placeholder isn't supposed to be accessed through imgur interface. Only if you type random letters you'll see it.

When you get big enough monitoring and integration testing start to look pretty similar. The most reliable kind of monitoring is the kind that actually uses your system, end-to-end.

On that topic, never name your child "TEST" if you ever plan to take them to the doctor or hospital...

The 3.2 million number is highly suggestive:

    3263847/(365*6*24)
    = 62

It sounds like they've been running an integration test which uploads an image once per minute for pretty much their entire existence (they're a little over 6 years old.)

This is sensible. It's the only way to get instant feedback when the site stops working for any reason.

The sampling was done only for the old-style 5 character URLs. They've since moved to 7 character URLs. Depending on how that 3.2 million number was reached, you maybe should be calculating rate using just the time they were at 5 character URLs. I don't know when the switch occurred though.

so that you can test production

That was a reference to this channel YouTube uses for testing: https://www.youtube.com/channel/UCsLiV4WJfkTEHH0b9PmRklw

So it's the same as the twist in the movie Amelie...

Life imitates art!

Also this answer from Imgur (via same thread):

"@nourishedcloud if we told you, we'd have to kill you"

https://twitter.com/imgurAPI/status/563792055651270656

Definitely some test image.

We do this on our CDN stuff as well which pokes it all on S3.

Our test image is creepy as fuck: http://i.imgur.com/GiTqw00.jpg

That's a pretty funny parody of BBC Test Card F: https://en.wikipedia.org/wiki/Test_Card_F

Yep that's the one. It was done by Noel Fielding and Julian Barratt under the guise of The Mighty Boosh.

Which you may recognize as the show behind Old Gregg

Is Old Gregg otherwise famous separate from the show?

In the US, most people just know the viral video of it and don't realize it's actually from a show

At least here in the US, Old Gregg is pretty well known outside of The Mighty Boosh. It went around independently as a viral video a few years ago and wasn't directly associated with The Mighty Boosh.

I'm in the US, and that's why I mentioned it... because it's so popular on its own, not many people realize it's from the Mighty Boosh or any show for that matter.

For some reason, I was unable to reply to petercooper's comment earlier. This was meant for him.

A "few"... it's almost 10 years old.

Please don't remind me of the quick and inevitable passing of time towards nothingness. That's a disrespectful thing to do on a Friday morning.

I remember at school the TV was wheeled in and the class was told we were going to watch something. Whatever the programme was, it wasn't on and we got the test card. The headmaster popped his head in and told us all to watch it carefully as the girl will draw on the board at the end.

We all watched it for what felt like hours. Strangest day ever.

A well communicative dummy image.

At a former job, our standard test image was the jellollama GIF.

From the Forbes article: In the longer term, Facebook Credits could become truly disruptive by becoming a currency for peer to peer lending, microtransactions, and for usage by the unbanked in emerging markets around the world.

It's amazing how much utter tripe gets written and published about this sort of stuff.

This chart is hard to read: http://i.imgur.com/HJYmCLN.png

Since the values are percentages, it's weird that the scale is out of 0.9 -- why not 100? Why leave out the other ~98% from the visualization? The radial arrangement makes it even harder to read and compare the values.

The effect (at least for me) was to make it seem like these images appeared much more often than they actually did. I actually found the raw numeric table below more illustrative!

Pie charts are already a terrible visualization, this is way worse. You're supposed to compare the area of these half circle shaped rotated slices easily? Should have just used a bar chart, in fact why wouldn't you use a bar chart? You could just have a single box and illustrate the percentage amounts by dividing the box up into different colors to indicate the percentages of a whole, but this? Nope. Tufte would probably not be pleased.

Pie charts are excellent for showing parts of a whole, and very poor for any other type of visualization. Just because people are misusing a chart type doesn't make it fundamentally useless, regardless of what Tufte thinks.

At a scale of 100% I feel that 0.84% would be awfully few pixels. That said I agree that a bar chart might have been more appropriate but c'mon - pretty!

Yeah, this "radial bar graph" trend is really baffling. I suspect some charting software added it as a new chart type and everyone's trying it out. Hopefully it will die soon.

What are the chances that imgur is throwing this image as an exploit prevention technique? Maybe there were 3000+ times the author triggered the rate limit, or some other metric?

The next one called "White Line" appeared 2000+ times too, maybe doing the same job.

In early days of reCAPTCHA at CMU, Luis von Ahn did a similar thing. There are a lot of sweatshops to type in CAPTCHAs, they detected that, and instead of blocking them, they threw at them longer (sometimes entire sentences) CAPTCHAs to get their help digitizing text.

This is interesting and it seems plausible. However, at the beginning of the article the author said they stumbled upon this image multiple times by manually entering random 5 character filenames. If it were an exploit prevention feature, I would hope it would be more difficult to trigger

It's an easteregg? that you get when accessing an imgur album as an individual image. Because the original user is scraping he basically found 3.2M galleries.

It only happens with older albums that do not have /a in the URL.

//EDIT: source: https://www.reddit.com/r/programming/comments/3clho3/1_out_o...

Clicking the link below creates a new imgur URL:

http://imgur.com/upload?url=https%3A%2F%2Fnews.ycombinator.c...

Perhaps someone had included a URL like the one above in some wordpress template.

I like this theory.

This is an old story. The answer is: this is a test image used in integration tests of Imgur itself.

I'd bet money that it's their standard test image.

Good call. What else would you bet on?

Why is the author assuming that different URL ids necessarily translate to different images/uploads? The character based id could simply contain redundant parts, which favour certain images over others.

Seems 1:1

http://imgur.com/blog/2013/01/18/more-characters-in-filename...

What I find strange is the use of a hash of the image. There are other methods to detect whether images are visually the same while they have a different resolution, etc.

From what I remember is that you'd resize all images to a fingerprint of yxy (lets say 100x100 pixels), apply some filtering in order to normalize lighting and then you'd XOR 2 fingerprints in order to see how much they are similar.

That's more complicated, more likely to false-positive, and I doubt it would provide much more value. Very few people are resizing images and uploading a slightly different version; the overwhelming majority of the time, duplicate images are people uploading the exact same file, byte-for-byte.

I too noticed that imgur urls were short some time ago so I put some JS code together and let it run in background tab for some time and what I've noticed is that there is ever so often some images that appear from time to time. There are at least that coke/pepsi logo history image and then there is that man with big slash on his face.

Never seen that facebook man before for some reason.

If you have a hosting place I can share JS. But basically it just randomizes urls and if the resulting image is "if(e.naturalWidth == 161 && e.naturalHeight == 81)" it removes that image as that is the default size of the "image not found" image.

Is this your site?

Have you tried any https://en.wikipedia.org/wiki/Steganalysis?

They are comparing SHA256 hashes, so the images are literally identical (and don't just appear to be identical visually).

I wish there were more of these little experiments being done! What with just a little more time and resources, imagine what other weird quirky discoveries are out there.

I agree, I've seen lots of fun things like this on the internet. But good ideas are hard to come by. I did something like this a few years back, for favicons:

http://tech.arantius.com/favicon-survey

Google `qxapoiu`.

Though it was more fun back in 2002 :)

For those wondering, "TARINGA" is an argentinian forum full of piracy. I assume some of those images are uploaded by users to illustrate their posts.

Apparently this is the result of exercising a bug in imgur.

monopoly mystery solved, but what's up with DJ David?

Do they have some kind of data deduplication by the way?

Given how an image can be hashed with something like SHA256, it's kind of surprising that uploading the same image twice doesn't result in the same URL.

Yeah, my thoughts exactly.

Well, the image title would be different, and that's a significant difference. "Smiling guy" and "Justin Bieber at my sister's birthday party" are very different :-)

Also, the "author" and comments wouldn't want to be merged -- different comment streams for different audiences.

The entry for the image can be different in the database, but there is no need to actually store the file twice, isn't it?

Good point. I hadn't considered the social angle because I usually avoid reading comments on sites like that.

Yeah, I've stumbled upon this Monopoly image about 2 years ago, when I was developing a small JS-based website. It dynamically embedded random pics from image hosting sites like Imgur, and this Monopoly guy was everywhere.

I guessed that it must have been some sort of massive Facebook spam at some point in time, or something like this. After all, the guy waves dollars in your face in exchange for credentials.

There is a link of DB attached(https://mega.nz/#!WBg1zAIA!dg0g2Q0kDm1q6r1WBMPe0-nrP3wxvokeF...). Can anyone verify weather the images actually point to "The Monopoly Man" or is a test image.

I tried this, and the second random string I tried was the image in question:

http://imgur.com/52nu1

That was actually the first image I got when trying 5 random characters on an URL.

Probably some kind of production canary or integration test in their deployments.

weird project link https://www.reddit.com/message/compose/?to=TheGamble&subject...

Reply on news.ycombinator.com