I have come across a post whose file seems to be a slightly optimized version of the source's file (about 22 kb smaller). They are pixel perfect duplicates and neither file has metadata.
Would it be worth replacing so the MD5 matches the source? This post in particular is rather old so the current version may be more common online (thus more likely to be MD5 searched).
Topic: Should optimized images be replaced with the source's original?
Posted under General
I tried that many years ago (between a post from FA and an optimised one from IB) and got rejected.
The standing was that if there was no difference in quality (i.e., it was a "1:1 match"), then the one already posted on e621 stays.
The only time a replacement will be accepted is if it's actually different, e.g. a fixed version, less compression, larger, etc
There's a few edge cases where metadata only replacements would be accepted (like removing rotation data), but those are rare
Updated
If it's a 1:1 match, I do wonder what the motivation of replacing it would actually be. Visually it's exactly the same, and e621 doesn't particularly care about anything that exists outside of the image itself.
Matching an MD5sum isn't a particularly useful trait because there's probably dozens of variations of any given image on the Internet and a single byte difference is going to result in a completely different md5sum.
Optimizing images may help slightly with bandwidth, but if that's a valid reason for a replacement it would open up most of e621's images to be optimized... which is going to drastically increase the storage space required considering the original images and any successive replacements are kept. Apparently around 90% of users are served downscaled samples anyway, since that's the default setting.
Alright, thanks for the input!
I didn't consider it a high priority, I just noticed my own copy was larger than the one here and was a little surprised when they came back pixel-for-pixel duplicates.