Last modified: 2014-09-10 08:33:58 UTC
A user has claimed that the fact that our image thumbnails don't preserve EXIF data violates the GFDL, viz "Preserve all the copyright notices of the Document." We use ImageMagick -thumbnail option to create thumbnails. That command destroys all "image profiles" and removes the EXIF data. Since the -thumbnail option is optimised for creating small files, probably the easiest way to solve this would be to 1) extract EXIF from original, 2) use -thumbnail to create thumb, 3) attach extracted EXIF to new thumb. That seems like a lot of overhead though.
GFDL-relevant data is on the image description page, not embedded in the image file. We strip extra data for thumbnails because that is the only sensible thing to do, avoiding useless multiplication of file size.
Hi! I can't comment on the GFDL aspect. Just a few words on the processor overhead: copying a 500 byte comment from a 3MB picture to its thumbnail (using rdjpgcom/wrjpgcom from the libjpeg-progs Debian package) is 130 times quicker (on my desktop, ymmv) than making the thumbnail itself with "convert -thumbnail". Therefore, with less than 1%, we can hardly speak of processor overhead. As for the disk overhead: I don't think that copying all the EXIF data would be required. If there are copyright information associated with the picture, they are in the comment field. I bet the average comment size among the 1.5M Jpeg pictures on Commons, doesn't exceed 100 bytes. Not that much disk overhead.
See bug 18871