Following my proposed experiment in data destruction, a few kind readers tried it out and let me know what happened1. I’ve summarised the results below, to try and see if there’s any common pattern.
| Software | Format | Was recovery possible? |
|---|---|---|
| Apple Preview | JPEG | No (rendered image still shown and could be captured via screenshot)2 |
| GIMP | JPEG | Yes (with minor alterations to the data, likely within allowed limits for JPEG)2 |
| Imagemagick display | JPEG | Yes (result not binary-identical)3 |
| Ubuntu Image Viewer | JPEG | No4 |
| Ubuntu Document Viewer | Yes4 | |
| PDF reader | PDF from a browser, stay in a PDF reader after the browser closes but can’t be saved5 | |
| Word (Windows 95) | DOC (on a floppy!) | No (but re-inserting the floppy worked!)6 |
As far as I can tell from this data, there isn’t much of a pattern here. Broadly, the observed behaviour seems to depend on the software rather than the format, and ‘viewer’ style applications appear less likely to allow re-saving than ’editor’ apps (but the behaviour of the Ubuntu Document Viewer shows this is not a robust finding). All we can be sure of at this point is this: “It’s complicated”.
To find out what’s going on, we’ll need to look more closely at what happens when we open a file…
Thanks also to Nick Krabbenhöft for pointing out that I could have been a bit more careful about my original experiment, and that would have helped work out where the JPEG differences came from in the case of re-saving the image from GIMP. That said, I expect such minor differences are down to small variations in the implementation of the JPEG decompression scheme, as permitted by the standard. i.e. my final image is likely the no more different that the same original image might be when rendered by a different software application. ↩︎
See the original post ↩︎ ↩︎
Result from @archivalistic ↩︎ ↩︎
From @andrewjbtw ↩︎
Also from @andrewjbtw ↩︎