Upload sniping / Uploading with few to no tags

Well, after reading a bit through the Discord chat, it seems that there is a big frustration wave going on against upload sniping.
That means that there is an user who uploads an image pretty fast from Pixiv and is only adding a few tags to the image before uploading (for. example "touhou 1girl solo"). And it seems that there are some users who consider this upload behavior as disdainfully against other users who aren't that fast.
This has been discussed at least one time before in topic #8824. It honestly switched pretty fast to another topic, but it was discussed partially.
So the question is if it is still ok to upload with few tags, because there is quite some beef going on between users because of this and it is really annoying to me, because we can't reach concensus that way.
So I thought I'd create a forum topic about this uploading behavior. But one also has to say that image gets tagged afterwards by the uploader.

Personally I usually upload things with most of the general tags I can see easily off the bat and then once it's uploaded I spend more time tagging everything I can think of, but considering things I upload are usually days old it's not really much to do with sniping or such.

Since I'm not one actively on the "hunt" to upload things it's not really something I'm concerned with overall, but admittedly I'd feel pretty annoyed if I was uploading a picture, taking some time to write like 10 tags just for it to get sniped by "1girl touhou solo" and the real tags coming several minutes later. So I can see that being especially annoying to contributors who take the time to tag properly before upload and getting sniped by contributors who just put 2 tags to upload as fast as possible.

Hillside_Moose said 5 years ago:

Being the first person to post the newest Pixiv art is serious business.

So this still stands I guess.

In my opinion, this is extremely unimportant, as long as the images are properly tagged by the uploader in the end.

Type-kun said:

So this still stands I guess.

In my opinion, this is extremely unimportant, as long as the images are properly tagged by the uploader in the end.

To what extent does "properly tagged by the uploader in the end" mean? 5 minutes? An hour? A whole day? Some uploaders straight up forget about a post until several hours later and nobody cares to let them know if it just goes missed without a search for something like gentags:<5.

Personally, I think the standard has changed -- it actually does matter who uploads a post because it is, quite possibly, the single most recognized thing on your profile related to contribution that no other user doesn't note, aside from translation. It is how new users get promoted to have unlimited contributions when they deservedly shouldn't IMO, and the sentiment is echoed correctly that the easiest way to get a low deletion count and have high quality posts attributed to your name is just to minimal tag snipe. It's not integral to encouraging a healthy uploading atmosphere.

And I recall DakuTree back years ago set up a bot to automatically upload posts from certain artists seconds after they post their work on pixiv with some sort of minimal auto-detection tagging system, and he got negged for that. Of course, no one gave him permission to, but it was still well within his ability to -- he could even just have used it on his own account without admitting it was a bot, and claiming that he's just faster than anyone else. Anyone can do it with enough work.

But all that is, is a difference between machine and human work. We should give more credit to people who tag posts thoroughly, even when it isn't their post. One user also got recently negged over the fact that he said "tagging isn't his job", when it is everyone's. It's easy to find good posts, it's hard to be honest about it.

EDIT: On that note, since you can watch the upload listing, you could just steal someone else's tags for yourself too, so you 'look' like a good upload sniper.

This problem – at least it’s considered a problem by many users, which I can totally understand – seems to be coming up again and again. As fighting against it with the currently available features of Danbooru – like negative feedback – doesn’t seem to work, maybe we need some feature to defuse the issue.

My spontaneous idea was the ability to “reserve” an image (MD5 hash?) for a short amount of time to give the uploader time to properly add tags and notify other users that this image is currently being uploaded.

• Reservation time should be long enough to properly tag the image but not too long, to avoid blocking an image if the uploader abandons it.
• Abandoned reservations should count towards some limit (deducted from some slowly filling token bucket or something like the upload limit) to prevent griefing by reserving lots of images but never actually uploading them, just to piss other users off. Try that a few times and enjoy your zero limit.
• “Abandoned reservation” means that the user doesn’t upload the image at all. It should be fine if it gets uploaded after the reservation expired. For example, I rarely upload, so I never worry about sniping and I’m pretty slow at tagging.
• Using the upload form or the bookmark should have buttons/a button to either reserve this post (opt-in) or release an automatic reservation (opt-out) to avoid accidentally trashing your reservation limit.
• There should be a JavaScript timer on the upload page, showing you how long you have for tagging until your reservation expires.
• Other users trying to upload the same image should receive a message like “This image is currently being uploaded by another user.”

I guess seeing the message that someone else is already uploading the image you wanted to upload can be quite annoying, but I think it’s less annoying than spending some time on tagging and then seeing that you got sniped by a three-tag upload. Less time wasted, move on to another upload.

I currently see two possible problems with my idea:

• If the reservation time is too short, like maybe one minute, the uploader might only spend only that one minute on tagging, unlike now, where some users throw in three tags first and thoroughly tag later.
• If the second uploader leaves after seeing the message about someone else uploading the image already but the first uploader abandons the reservation, the image might not get uploaded at all. Maybe there could be an option for the second uploader like “Try to reserve this image after the current reservation expires” and then you leave the page open. Once the first reservation expires, either you get the reservation or get taken to the uploaded image. Might be a bit risky if you’re not watching that page.

This is a pretty big change and probably not easy to implement, but something like this might be the only option to soften this long-standing annoyance. Let’s face it: upload sniping isn’t going away otherwise, right?

Thoughts?

Mikaeri said:

EDIT: On that note, since you can watch the upload listing, you could just steal someone else's tags for yourself too, so you 'look' like a good upload sniper.

Urgh, thanks for reminding me of that user who kept taking tags from random other posts... that had little to do with the post he put them on.

Mikaeri said:

EDIT: On that note, since you can watch the upload listing, you could just steal someone else's tags for yourself too, so you 'look' like a good upload sniper.

I vaguely remember from the code that if you try to upload the image and somebody already did, the tags you've entered are supposed to get merged into the existing post under your name, as if you found it and edited the tags in. I'm not sure if it's the case now, would have to dive into the code, don't have time right now.

Also, we already have a report showing how well the user takes care of their uploads. This is enough to decide on their promotion.

Besides, you make it sound like uploading art from already established good quality artists is the privilege of users who are already granted unrestricted uploads. If regular user managed to do it first, even with minimal tags, then so be it. If they don't tag properly within certain time period, they should face repercussions, as well as higher-level users. As for how long is "certain tag period", I'd say up to two hours is fine.

kittey said:

My spontaneous idea was the ability to “reserve” an image (MD5 hash?) for a short amount of time to give the uploader time to properly add tags and notify other users that this image is currently being uploaded.

I've had a similar idea, but I felt that it creates more problems than it solves. Maybe worth a try, but I wouldn't bother personally.

Type-kun said:

I vaguely remember from the code that if you try to upload the image and somebody already did, the tags you've entered are supposed to get merged into the existing post under your name, as if you found it and edited the tags in. I'm not sure if it's the case now, would have to dive into the code, don't have time right now.

That would be really nice if that were possible.
Well, if I understood you correctly, then I've made a short test with the tag ":d". But it seems that it didn't got merged.

Type-kun said:

I vaguely remember from the code that if you try to upload the image and somebody already did, the tags you've entered are supposed to get merged into the existing post under your name, as if you found it and edited the tags in. I'm not sure if it's the case now, would have to dive into the code, don't have time right now.

I am pretty sure this is not how it works -- if this were actually so, then I'd see a lot more tag merging on posts from popular artists. Instead, some people just tag 1girl solo and call it a day (so that it shows that they edited basic tags on an upload snipe and registered that edit).

Also, we already have a report showing how well the user takes care of their uploads. This is enough to decide on their promotion.

Besides, you make it sound like uploading art from already established good quality artists is the privilege of users who are already granted unrestricted uploads. If regular user managed to do it first, even with minimal tags, then so be it. If they don't tag properly within certain time period, they should face repercussions, as well as higher-level users. As for how long is "certain tag period", I'd say up to two hours is fine.

Is that not the case? As approvers, we don't make it our job to look through all the posts that go through the gallery to see if they were properly tagged in a substantial time -- we really only look at posts that are pending, and we'll see disproportionately that regular users more frequently 1-3tag upload than that of users with unrestricted up perms. Because the fact of the matter is that when a post is poorly tagged and left poorly tagged for 2-4+ hours, and it's already active because it skipped the queue, then we notice it less, less than if it were in the blue queue and we were obliged to double-check that.

So that's why I find it pretty funny -- you get promoted to unrestricted uploads, and boom! Now you no longer have to be so diligent to take care of your tags because we, the approver team, are less likely to notice your poor tagging and thus less likely to neutral or negative feedback you on it. There are a handful of users that I think don't deserve unrestricted uploads, but trying to demote them is a pretty arduous effort that will be scrutinized by a handful of the other upper staff. But if what you're saying is right then it should be considered.

I've had a similar idea, but I felt that it creates more problems than it solves. Maybe worth a try, but I wouldn't bother personally.

I think kittey's idea is worth an implementation. In my opinion, if the site should continue to grow, it should foster a more collaborative environment rather than a competitive (and lazy) one.

Type-kun said:

I vaguely remember from the code that if you try to upload the image and somebody already did, the tags you've entered are supposed to get merged into the existing post under your name, as if you found it and edited the tags in. I'm not sure if it's the case now, would have to dive into the code, don't have time right now.

This used to be the case some years ago, when trying to upload an image that was already uploaded would take you directly to the post instead of displaying a generic "This upload is a duplicate" message like it does now.

Uploading is like capitalism. The end-user doesn't care who uploaded the image or how, as long as they get the product. People who tag up-front are small business owners who will never amount to more, while min-taggers are corporations, who believe the ends justify the means.

People with integrity will always stand at a disadvantage to those who will skirt or bend the rules, but regardless of this, I'll still continue trucking on as I have been because I don't want to stoop to their level.

kittey said:

My spontaneous idea was the ability to “reserve” an image (MD5 hash?)...

Type-kun said:

I've had a similar idea, but I felt that it creates more problems than it solves. Maybe worth a try, but I wouldn't bother personally.

This would make uploading a two-step process: first you upload the file, then you fill out the metadata and submit the upload. This would help with certain other problems:

• Would be able to validate the file and perform dupe checks earlier, before the user wastes time tagging it.
• Image previewing would be easier (wouldn't need to proxy hotlinked images through our image proxy).
• Wouldn't need to download the file multiple times (currently we download it twice, once by Danbooru and once by IQDB).
• Would be able to generate samples/thumbnails in the background while the user is tagging.
• Would be able to determine the image's filesize and resolution less hackishly (currently we do a HEAD request to get the Content-Length, but reportedly this fails sometimes).
• Uploading would be a lot faster if you could just hit ctrl+V then start tagging, like you can on Imgur.

On the other hand, a full reservation system would add a fair amount of complexity. So I think it's not a bad idea, but it would take quite some effort to implement.

CodeKyuubi said:

Uploading is like capitalism. The end-user doesn't care who uploaded the image or how, as long as they get the product.

Pretty much. I don't give a shit about who gets first post or who stole whose upload. I do care about uploaders making the rest of us pick up the slack on their tagging though.

I think we've brought this on ourselves via a perverse system of incentives. The only thing you get recognition for on Danbooru is uploading, which incentives certain people to ignore tagging in favor of getting their name on as many uploads as humanly possible.

The last paragraph is not correct imo. I think this becomes clear when you look at who is actually feedback and for what.
Moonspeaker, henmere, Paracite are getting a lot feedbacks ("recognition") because of their work.
Then you, BrokenEagle, Toks and RaisingK got theit feedbacks because of the technical stuff you do here.
Then iridescent slime has no uploads at all and is a prolific tagger and good in forum discussions.

So I am really in favor to say that uploading is getting the least recognition of all user activity, next to approving posts (if you weren't Not One Of Us who looked that images don't get deleted at day three).

evazion said:

This would make uploading a two-step process: first you upload the file, then you fill out the metadata and submit the upload. This would help with certain other problems:

• Would be able to validate the file and perform dupe checks earlier, before the user wastes time tagging it.
• Image previewing would be easier (wouldn't need to proxy hotlinked images through our image proxy).
• Wouldn't need to download the file multiple times (currently we download it twice, once by Danbooru and once by IQDB).
• Would be able to generate samples/thumbnails in the background while the user is tagging.
• Would be able to determine the image's filesize and resolution less hackishly (currently we do a HEAD request to get the Content-Length, but reportedly this fails sometimes).
• Uploading would be a lot faster if you could just hit ctrl+V then start tagging, like you can on Imgur.

On the other hand, a full reservation system would add a fair amount of complexity. So I think it's not a bad idea, but it would take quite some effort to implement.

This would suffer from the same situation as naive reserving - when user uploaded the file but abandoned it midway for whatever reason. The timeout for "reserving" should be reasonable for this not to happen, something about 5 minutes, but full implementation would also require to delete/reuse the previously uploaded files.

Also, if implemented, one user should be able to reserve one md5 at time; otherwise we'll surely end up with people opening 20 tabs and hastily tagging through them all within reservation window.

The problem I see with the reservation system is that it probably won't change anything.
So, you are having reserved the posts and tag it afterwards. According to howto:tag_checklist the tagging threshol is extremely low, though. So you can reserve the post, add 6-9 more tags and say that's it, because it is still acceptable tagging and then go to the next upload.
I'm likely to say that this might even encourage even worse tagging.

CodeKyuubi said:

Uploading is like capitalism. The end-user doesn't care who uploaded the image or how, as long as they get the product. People who tag up-front are small business owners who will never amount to more, while min-taggers are corporations, who believe the ends justify the means.

People with integrity will always stand at a disadvantage to those who will skirt or bend the rules, but regardless of this, I'll still continue trucking on as I have been because I don't want to stoop to their level.

evazion said:

I think we've brought this on ourselves via a perverse system of incentives. The only thing you get recognition for on Danbooru is uploading, which incentives certain people to ignore tagging in favor of getting their name on as many uploads as humanly possible.

Amen.

Provence. said:

The last paragraph is not correct imo. I think this becomes clear when you look at who is actually feedback and for what.
Moonspeaker, henmere, Paracite are getting a lot feedbacks ("recognition") because of their work.
Then you, BrokenEagle, Toks and RaisingK got theit feedbacks because of the technical stuff you do here.
Then iridescent slime has no uploads at all and is a prolific tagger and good in forum discussions.

So I am really in favor to say that uploading is getting the least recognition of all user activity, next to approving posts (if you weren't Not One Of Us who looked that images don't get deleted at day three).

Alright, this is where your naiveté is really showing. Yes, they get a lot of feedback, but those are from users that have worked very long and hard to keep at what they've been doing for the past 5 or 6 years. People praise them because its easy to praise their work, but that doesn't necessarily translate to anything tangibly feel-good. What really gets people happy and recognized is if good content gets uploaded under their name. It is the one thing there that sits next to a post out in the open. If you want to see who tagged a post, who translated a post, who provided commentary, you have to dig through that edit history. Some users will recognize that, and that is excellent. But uploading? It's a nice little trophy to wiggle that e-peen around a bit. And then you have a few people who will more frequently browse your posts for good uploads, regardless of how fairly (or moreso, unfairly) you got them.

This isn't activity that should be rewarded either, just 'finding' the most popular posts by the most popular artists and then crapping them out the gate 3-5 seconds right after posting with minimal tags. Giving feedbacks for uploads have went to crap because of the perception of positively repping upload snipers. That's not something you or anyone should promote. I positive feedback users that upload content that otherwise would've gone missed (the likes of iphn or kudus), not the kind of stuff I could find by just browsing and filtering my feed by whatever got 500+ bookmarks on pixiv.

evazion said:

This would make uploading a two-step process: first you upload the file, then you fill out the metadata and submit the upload. This would help with certain other problems:

• Would be able to validate the file and perform dupe checks earlier, before the user wastes time tagging it.
• Image previewing would be easier (wouldn't need to proxy hotlinked images through our image proxy).
• Wouldn't need to download the file multiple times (currently we download it twice, once by Danbooru and once by IQDB).
• Would be able to generate samples/thumbnails in the background while the user is tagging.
• Would be able to determine the image's filesize and resolution less hackishly (currently we do a HEAD request to get the Content-Length, but reportedly this fails sometimes).
• Uploading would be a lot faster if you could just hit ctrl+V then start tagging, like you can on Imgur.

On the other hand, a full reservation system would add a fair amount of complexity. So I think it's not a bad idea, but it would take quite some effort to implement.

Type-kun said:

This would suffer from the same situation as naive reserving - when user uploaded the file but abandoned it midway for whatever reason. The timeout for "reserving" should be reasonable for this not to happen, something about 5 minutes, but full implementation would also require to delete/reuse the previously uploaded files.

Also, if implemented, one user should be able to reserve one md5 at time; otherwise we'll surely end up with people opening 20 tabs and hastily tagging through them all within reservation window.

I'm willing to give it a shot and help debug the new feature if it comes to fruition. Something as advanced as this will be privy to a number of bugs, but I think if we can manage to accomplish this it will bolster the number of active contributors that are diligent about having work be well-tagged on the site.

Still though, just one concern I have about this proposition is the fact that if you reserve a post, you should be bound by some restriction to tag it truly correct or suffer some blowback (not being able to reserve posts at another time if other users have to correct a ton of your tags, or you bogus tag and up it just to fix it seconds later), but I'm assuming those details will start ironing out when progress begins on it.

You really think that? Well, I can only say that this apparant background work which it probably isn't, still receives a lot of praise and while it isn't shown to you directly, those "fairies" at work are also what also keeps Danbooru running.
Maybe you should actually keep looking a bit further than only the uploads; the whole thing you consider behind the scenes is probably what users also like: Artist gardening, translating, tag gardening, technical stuff. I mean, this is an amalgamation of work and uploading takes only one part in it: Uploading is up to anyone, translating for example not.

Well, I just hope that this maybe upcoming feature might actually help in the end and that is the more relevant topic.

Isn't it just easier when you upload a pic, it has to have at least number of tags. Say like ten. and get negative rep for tagging with crap/irrelevant tags

Maybe it's because I'm more of an upload scavenger than an upload sniper but is this really a big enough problem to change the way we've been uploading stuff for over a decade now? Like Type-kun said, this is extremely unimportant.