[SkycladGuardian]

http://www.dkriesel.com/en/blog/2013/0802_xerox-workcentres_are_switching_written_numbers_when_scanning

Though it is old news (the bug was discovered a year ago) I've only learned about it recently and I'm still staggered by the implications. I mean, Xerox could be sued to bankruptcy if one could prove that damages were caused by this bug.

For those too lazy to click the link: A bug in Xerox' compression algorithms for professional scanners and photocopiers caused the device to switch written numbers on scanned documents, potentially altering important data. Oh, and this had gone on for eight years... 

[Aardwolf]

Compression algorithms? Why would they even... Huh?

[Phantom Hoover]

The article makes some attempt to explain why this happened near the end, but the quick version is that the compression algorithm they used saves space by using one patch of the document to fill in another if they look sufficiently similar. It looks like it's just been horrendously misapplied here.

[SkycladGuardian]

Yeah they used pattern matching to save memory but it was configured a little bit too imprecise...

[Klaustrophobia]

ya know, when I COPY something, I'm not looking for similar

[Phantom Hoover]

Yes you are, have you really never seen a ****ty but usable photocopy? Xerox's fatal mistake here was using the wrong kind of 'similar'.

[jr2]

Why didn't they just compress to JPEG or similar?  That would have worked (3-5MB -> 300-700 KB), and not done this...  :headdesk:

[Dark RevenantX]

Why didn't they just compress to JPEG or similar?  That would have worked (3-5MB -> 300-700 KB), and not done this...  :headdesk:

They probably are using JPEG-based compression to begin with, just adding this crap as an additional layer to improve PSNR.

[Aardwolf]

I, for one, had to Google that abbreviation. "Peak Signal-to-Noise Ratio".

(I knew SNR, but not PSNR)