Slashdot Mirror
Understanding the Antikythera Mechanism (hackaday.com)
szczys writes: We attribute great thinking to ancient Greece. This is exemplified by the Antikythera Mechanism. Fragments of the mechanism were found in a shipwreck first discovered in 1900 and visited by researchers several times over the next century. It is believed to be a method of tracking the calendar and is the first known example of what are now common-yet-complicated engineering mechanisms like the differential gear. A few working reproductions have been produced and make it clear that whomever designed this had an advanced understanding of complex gear ratios and their ability to track the passage of time and celestial bodies. Last year research by two scientists suggested that the device might be much older than previously thought.
Oh, I didn't mean to imply that it didn't, just that you have a better chance of recovering whatever it was by getting a detailed image of the magnetic domains than simply dropping the disk/tape into a (hopefully expertly restored) compatible machine. If certain properties of the recording system are known, such as it using a certain run length code that is DC balanced, then when a value is present that violates the allowable symbols on the disk, then the magnetic (and hopefully an aligned optical) image of the surface can reveal which bits are likely corrupted, and make a manual restoration.
The analogy to this is a printed manuscript, vs. a photocopy, vs. an OCR. If you have the OCR, then the only knowledge you have is that there was an error (or even less) in transcribing, if you can go back to the photocopy, which is analagous to an analog recording from the devices read head, then you can make a probalistic estimate as to which bit(s) were in error to result in the corrupted checksum, but the most definitive answer comes from a careful study of the original media magnetic and visible surface.
For example, say you have a CRC error on an old zip file from a DOS format floppy: if you can go back to the original media, and see in the optical domain that there is a scratch across the surface, and register that against the magnetic domains on the disk, then you can make a probable guess that the bits that are corrupt are the ones under the scratch, and try flipping them until you get a correct CRC. If it is a CRC32, then there is a very low probability that flipping those bits in an incorrect manner results in a valid checksum, and so you can assume that the set of bit flips that results in a correct sum when you attempt to flip those that fall under the scratch (or hunk of mold, which should be delicately cleaned off too), is the correct original data. If you were just looking at a binary dump of the media, then you would have no information to guide your guess work, and if you were to flip bits entirely at random, there is a lower chance that you can recover the original data, and a low confidence that you made the right correction (though if it's ASCII/EBCDIC/etc text, it's quite easy to use common sense to recover sparse bit errors).
Overall, though, I expect the century of B&W photography to be the best documented, by virtue of that technology's inherent stability.
Of course. But I've also found laser printed documents appear to survive quite well, with the limit imposed by the paper rather than the toner. I imagine laser printing onto polymer sheets like OHP film would result in excellent longterm survivability for important documents. There are companies that will record data onto gold plates with electron beam or focused ion beam recording techniques, as long as the chain of custody is properly maintained, these artefacts would be reasonably assumed to survive a millenia and still be readable.