The tag alias #80907 male_human/female_anthro -> human_on_anthro is pending approval.
Reason: Same reasoning as alias #70350. Just search human_on_anthro male_human female_anthro.
Topic: Tag alias: male_human/female_anthro -> human_on_anthro
Posted under Tag Alias and Implication Suggestions
A question arises - I've seen this kind of change made several times, and it inherently loses information that might be of interest to someone searching.
(Arbitrary example: human_male having sex with anthro_male, while anthro_female is present in the image. "human_on_anthro male_human female_anthro" catches that image, despite there not existing any direct male_human/female_anthro activity in the image. https://e621.net/posts/5700393 is an example, caught by the proposed replacement search, that male_human/female_anthro is not an accurate descriptor for.)
It's more literally accurate that this (and the many similar cases) are an implication and not an alias: "male_human/female_anthro" implies "human_on_anthro" (as well as "male_human","female_anthro","male/female"; which themselves technically imply "male", "human", "female", and "anthro".)
So the question, generically, is "why perform this kind of lossy aliasing"?
I can think of several potential reasons:
Processing speed? (Implications aren't free to process on the backend, and suffer from combinatorial explosion as seen above, whereas aliases can be replaced in the search query before the search even happens.)
Curation of tagging style? (For the same reason that the proposed replacement is "human_on_anthro" instead of "human/anthro", there is a style of tags that tends to get conformed to, over time, which helps people more easily guess tags they don't specifically know beforehand.)
I was therefore just interested in whether there existed a canonical answer for why these kinds of replacements should be done, as I'm not particularly familiar with the discourse around them.
(Largely-irrelevant pre-context not wholly relevant to the discussion: this originally came to my attention with the tag "maned_lioness" which long ago was aliased into "lion" and otherwise deleted; later "reverse_sexual_dimorphism" was added as a tag to cover that kind of situation (allowing for "female lion reverse_sexual_dimorphism"), but that similarly doesn't cover the generic case of lion-of-any-gender and female-peafowl-with-stereotypically-male-plumage in the same image, rare though that case may be.
It's the same generic problem of tag clusters being applied to a single character in an image - there's no effective way to search for "a male canine anthro and a female feral feline" in the same image (with "male canine anthro female feline feral") without getting male felines/ferals and female canines/anthros (as well as the other half of the combinations) as well. There's possibilities for how to encode such things into a search ("anthro&canine&male feline&female&feral" comes to mind, letting you specify groups of tags that must all be applied together), but that'd be a major design change to how the tagging system works in general (the same tag can occur multiple times on a single image - you might have "male&canine female&canine female&feline" (but not "male&feline") tag clusters on a single image, which makes tagging and searching yet_more_complicated) and would require basically every image to be incrementally retagged over time, so I dismiss that out-of-hand as an option. (It would however allow the search "human_on_anthro&(male&human)&(female&anthro)" to end up with the exact intended results, after sufficient retagging, so it's a pleasant idea at the very least.))
ixionkrystis said:
A question arises - I've seen this kind of change made several times, and it inherently loses information that might be of interest to someone searching.(Arbitrary example: human_male having sex with anthro_male, while anthro_female is present in the image. "human_on_anthro male_human female_anthro" catches that image, despite there not existing any direct male_human/female_anthro activity in the image. https://e621.net/posts/5700393 is an example, caught by the proposed replacement search, that male_human/female_anthro is not an accurate descriptor for.)
It's more literally accurate that this (and the many similar cases) are an implication and not an alias: "male_human/female_anthro" implies "human_on_anthro" (as well as "male_human","female_anthro","male/female"; which themselves technically imply "male", "human", "female", and "anthro".)
So the question, generically, is "why perform this kind of lossy aliasing"?
I can think of several potential reasons:
Processing speed? (Implications aren't free to process on the backend, and suffer from combinatorial explosion as seen above, whereas aliases can be replaced in the search query before the search even happens.)
Curation of tagging style? (For the same reason that the proposed replacement is "human_on_anthro" instead of "human/anthro", there is a style of tags that tends to get conformed to, over time, which helps people more easily guess tags they don't specifically know beforehand.)
I was therefore just interested in whether there existed a canonical answer for why these kinds of replacements should be done, as I'm not particularly familiar with the discourse around them.
(Largely-irrelevant pre-context not wholly relevant to the discussion: this originally came to my attention with the tag "maned_lioness" which long ago was aliased into "lion" and otherwise deleted; later "reverse_sexual_dimorphism" was added as a tag to cover that kind of situation (allowing for "female lion reverse_sexual_dimorphism"), but that similarly doesn't cover the generic case of lion-of-any-gender and female-peafowl-with-stereotypically-male-plumage in the same image, rare though that case may be.
It's the same generic problem of tag clusters being applied to a single character in an image - there's no effective way to search for "a male canine anthro and a female feral feline" in the same image (with "male canine anthro female feline feral") without getting male felines/ferals and female canines/anthros (as well as the other half of the combinations) as well. There's possibilities for how to encode such things into a search ("anthro&canine&male feline&female&feral" comes to mind, letting you specify groups of tags that must all be applied together), but that'd be a major design change to how the tagging system works in general (the same tag can occur multiple times on a single image - you might have "male&canine female&canine female&feline" (but not "male&feline") tag clusters on a single image, which makes tagging and searching yet_more_complicated) and would require basically every image to be incrementally retagged over time, so I dismiss that out-of-hand as an option. (It would however allow the search "human_on_anthro&(male&human)&(female&anthro)" to end up with the exact intended results, after sufficient retagging, so it's a pleasant idea at the very least.))
This just means the OP's query is missing male/female.
The short version is that there is simply a level of specificity in a tag that is just too much when maintenance and implications are required.
lafcadio said:
Oh no, we couldn't just use pregnant and krystal in combination, no, we have to use pregnant_krystal.
I won't say it's a slippery slope, because we can determine an acceptable level of specificity whenever we want, and some special exceptions like wolf_link look overspecific but are not (see wolf_link -wolf), but any person who could tag male_human/female_anthro could very well just tag male_human female_anthro male/female human_on_anthro instead, and when the individual parts are well-supported then one can also use tag grouping to, say, find male_human ( ~female_anthro ~female_feral ) male/female ( ~human_on_anthro ~human_on_feral ) instead of then creating male_human/female_feral and requiring further implications + maintenance.
Thanks for the quick reply; basically as I assumed, "a somewhat-arbitrary (but deliberate) collective executive decision as to the depth of acceptable specificity in compound tags".
That's all I particularly wanted - thank you.
lafcadio said:
This just means the OP's query is missing male/female.
As to this specific linked counterexample, it's true that one can include the additional "male/female" tag and get closer to the intended purpose, but "human_on_anthro male_human female_anthro male/female" would capture a single image containing two scenes of a human male/anthro male pairing and an anthro male/anthro female pairing, despite lacking the specific situation this covers.
To clarify, that's fine. In any of these compound-implication aliases, there's some pathological counterexample hiding somewhere; but that's true of every alias, so it's inherently fine - aliasing all existing "panties_pulled_aside" tags to the at-the-time more-common "panties_aside", for example, lost the specific "pulled" aspect of the original sub-tag (makes it harder to distinguish between a deliberate action on the part of the wearer vs a wardrobe malfunction). There's no escaping that; it's just the nature of the beast when making an alias of any kind.
(The proposal of the &operator was more of an aspirational perfect-case thought - I'm aware it's not supported in the syntax and would get naively parsed as a new tag all its own, if it's not even an outright illegal character in the tagging field (I haven't checked). Was primarily a solution to the "human_male" vs "male_human" arbitrary-canonical-representation problem - using a new delimiter to allow both of those to be interchangeable without having to manually manage aliases and implications - whether tagging or searching for "human&male" or "male&human", being able to identify that those are intended to be a paired property of a single entity.)
Regardless of my further digression; thanks again for the response.