I can rescan old contests, but we don’t want to convert those links to the cached version because the cache URLs change periodically (I think I read somewhere that it was every 15 minutes) specifically to prevent hotlinking. That’s why clicking on a thumbnail on Bibliostats from contests #418 forward (except for the active contest) takes you to a Fark 403/forbidden page – the URL for that image that was scanned at contest close has changed and is no longer valid.
So, if I replace the existing URLs in old contests with the cached versions, all pop-to-original will stop working, even for older contests where it mostly works now (except where the original has been deleted).
But that did spur a thought on the suggestion of using hashes. Elsinore, can you ask Mike if images are re-saved to jpg when being cached or if they’re copied as is? In other words, when Fark caches in image, does it cache the exact image from its source or does it use an imaging library to reconvert the uncompressed image to jpg?
If it re-converts to jpg or otherwise does anything to alter the image file, then the hash for the cached version wouldn’t match the hash for a repost of the original image and thus no scarlet R checking. That’s the nature of hashes – they’re designed to produce significantly different outputs in response to tiny changes in input.
