Type-kun said:
Hmm. I wonder how feasible would it be to implement an automatic image comparison using imagemagick, and require a second approval if there's something significantly different between the two.
Would be in favor of it.
I'm flipping my stance on this by the way. Given the way post #2637690 (and the subsequent parent) turned out, I don't think it's fair that we start just replacing images out of the blue even when we know one might be considerably lower quality than the other.
Reasoning for this is that even though as approvers we can and are allowed to do it, it feels especially murky to begin with, that we are allowed to handle this the way we do. In a case like that either one has to take an archiver's stance (in which case, all versions of a post, even if inferior, are 'saved') or a minimalist's stance (in which case posts must always attempt to be replaced with their most known superior counterpart).
The first I can't really find any fault with. We've been doing it that way for years before the post replacement feature came out due to NWF Renim's request after a certain topic came to light, about flagging samples without their originals. And yet we're considering using it for alternate purposes that might be unjustified to some, and might even confuse users that are doing things perfectly fine, thinking that they now have to consider posting in topic #14156 when they shouldn't have to.
Look, duplicates suck, I know. No one wants them, but for some of these we can't even consider to be 'duplicates' anyway. post #2637690 is a jpg that is much more lightweight than its png parent. Some people like that. Some people also like having the Twitter source be easily linked to share in the future. Some people like smaller images because it's less space on their computer or easier to view on their devices. We can't cater to all of those preferences if we take on a minimalist stance where all the inferior posts that should be replaced will be replaced. It's confusing and prone to mistakes, especially if a final post on pixiv/seiga is revised as I mentioned in forum #132898. Example is post #2750489 and post #2749186.
The replacement feature is something extremely powerful that is currently only available on our booru. It also creates potential headaches for other boorus and services in general (IQDB and SauceNAO) when they index posts that are actually different than what's live on our server. If part of the appeal in curation is to also expose more users to an artist's work, then their failure is also our failure.
Anyways, let me outline what I'm completely fine with using the replacement feature for before I forget.
- Sample post -> Non-sampled post. Original intent for the feature.
- Same image, but file with stripped or inaccurate/corrupt metadata -> better image. post #2196287 is one such example.
- Same image, but worse compression -> Same image with better compression. post #2766377
- Sample post with already upped original -> original post from alternate source. Less deletions on an artist's record where preventable is a + for me. post #2753690, post #2746728, post #2763253
What I am against:
- Inferior Twitter (or other source) jpg -> superior pixiv/nicoseiga/nijie png. Some people might prefer the 'inferior' post, as I mentioned above.
- Unsourced lossy conversion (third party sample) or lossy-lossless conversion -> original image. We should not credit uploaders that find posts from alternate, untrusted sources.
- Posts with clear visual changes due to revision or WIP -> Finished/revised posts. For obvious reasons.
Pinging @Randeel on this because I know we had beef over this issue, along with @NWF_Renim. I think we need a more ground stance on this, so as to not lead uploaders and approvers alike to confusion.