antinegationism

[antinegationism]

I was weak and a coward and it only freed up 8%. This process always finds some new way to fuck with me.

Deleting blogs is extremely stressful each and every time but I fear intimacy and refuse any form of emotional support. Don't talk to me.

[antinegationism]

Deleting blogs is extremely stressful each and every time but I fear intimacy and refuse any form of emotional support. Don't talk to me.

[antinegationism]

Man you guys don't even know the half of how weird these disincentives can get. I mean, on one level you probably know exactly how weird they can get, in that if I posit any hypothetical, you will get the right answer by just blurting out your first gut feeling and then not thinking the slightest bit beyond that. But on the other hand you as a single person are not enough people to posit the space of hypotheticals.

It's odd to me that audience capture hasn't yet entered the general Discourse lexicon the way phrases like "motte-and-bailey", "virtue signaling", etc. have. For my money it's the main dynamic that drives nearly all sufficiently-famous Internet polemicists into crazy wingnuttery; by extension it's one of the most powerful forces on the Internet and mass culture in general, and should probably be more carefully studied.

[antinegationism]

when will siikr return from the war

Have you ever considered that the reason you are so interested in finding old posts, is because your new ones just aren't good enough?

[antinegationism]

Second this.

I read the series completely divorced from the fandom and enjoyed it a lot. And this was WHILE the fandom was a thing that was happening.

One time I even encountered someone from the fandom and they were just completely shocked that someone was just like, out there enjoying it privately and alone, and I was shocked that they were shocked.

I still don't know which of us was wrong but it couldn't have been me.

On the one hand roughly my entire online social circle and also a bizarre number of the creators of my favorite recent media are fans of it, so objectively I probably really should try reading Homestuck at some point.

But on the other I am incredibly stubborn spiteful person and at this point 'never read homestuck' is, like, a core part of my identity I think.

[antinegationism]

siikr is giving a "database is starting up" error when trying to search a blog btw

it's lying. The database is just too embarrassed to admit that it has become too fat to pick itself up off the couch.

[antinegationism]

That post is going around again, so, POWER OF BRANDING time: there are two (and only two) places. There is the Town Square Where All The Good And Normal People Do Their Business. And there is the Evil Cave On The Outskirts Of Town Where No One Is Supposed To Go. If at any given time there are more people in the Evil Cave than in the Town Square, the Town Square will magically blow up and everyone there will die. Is the rational and self-interested move to transfer your business immediately from the Town Square to the Evil Cave?

Absolutely not. By definition there are more potential customers in the town square than there are in the cave, right up until the moment when there are 0 customers in the town square.

The rational and self interested move is to open up a second branch in the cave, and staff it in proportion to the population of the cave, which would ideally remain 0, because it's a cave and how the heck is it managing to attract so many people to begin with? Note however, that this scenario is not isomorphic to the pill or waiting room dilemmas. Since in this scenario, everyone can only be in the town or the cave, and they all start in the town. Which means that unless everyone hits a button to teleport to the cave at the same exact time, then the last 50% of people to hit the button die. There needs to be a third neutral place everyone starts but cannot remain in.

[antinegationism]

Blog deletion roulette is coming up again soon.

Siikr now features a minigame.

It keeps score.

Are these facts related? One does wonder . . .

[antinegationism]

the catgirls are a nice and very welcome touch to siikr. your work on this website has been invaluable. do you have a ko-fi or something similar?

Yes thank you. It was a lot of work but I feel they are critical to site functionality.

I don't have a ko-fi but am actively interested in crowdsourcing some disk space to not have to keep deleting blogs every few months.

[antinegationism]

Answer is 3. It is both the smallest option. And maximally non-obviously composite by virtue of not even being composite.

[antinegationism]

Oh wow, y'all suddenly got real quiet real fast, huh?

Hey. Can you spin up a database on your local machine with 30'000'000'000 posts with perhaps ten times more tags and see how fast your search is. Please expect at least 1000 qps too.

(Context: Anon is (very stupidly) trying to poke fun at this post) Siikr currently indexes 44,325,274 posts. Due to my own implementation laziness, it does so extremely inefficiently, with each post storing a lexemetized english GIN index and language agnostic GIN index, and each post additionally containing its rawtext, json blocks, and html content representation. Furthermore, it does this with MASSIVE redundancy, because most posts are reblogs of other posts, but siikr treats them all as if they were text instances in their own right. Despite all of that inefficiency, it currently takes a grand total of 136GB of storage, running on a single virtual machine with a single virtual core and 1GB of (mostly unused) RAM.

44,325,274 posts is about 0.15% of 30,000,000,000.

If I were to instead host siikr on my old laptop (instead of the cheapest instance Digital Ocean is willing to sell me), then without making any code changes, it could easily accommodate 2% of the provided figure. If I were to modify the code to be smart about reblogs and avoid redundant representations, I'd venture to say maybe 5-15% of the figure. Which means I'd need about 10-20 laptops for siikr to host 30,000,000,000 posts. I don't know how many posts tumblr actually handles, but 30 billion seems on the low end. So probably you would need maybe an order of magnitude or two more laptops. (Note, that even this is with absolutely minimal optimizations, and I suspect that if I really wanted to I could in fact get 30 billion posts into a single laptop, first by switching to binary json or flatpack instead of raw utf8 json strings, then second by compressing post-content, and only decompressing it client side.) (Note also that this is for SEARCH. Which is a much harder problem than literally just showing people posts by users they're following, in that the latter is way more amenable to routing, cache optimizations due to the nature of viral content, decentralized storage, and compression/archiving of rarely accessed content.) On a broader point, all of the commentary on the operating costs of tumblr is friggen insane.

Like, I don't wanna be mean but, has everyone gone friggen braindead? Have we as a species forgotten that servers used to be a thing we were capable of managing prior to Amazon and Google overcharging you for them? If you force all videos, images, and audio to be externally hosted and linked to, the operating costs are stupidly cheap. Possibly cheap enough that --should there not be enough kind souls willing to leave an old laptop plugged in -- you could pay for maintenance by charging users for image and video hosting. Now, to be absolutely clear, I have no actual interest in replicating or managing a semi-decentralized community maintained version of this god forsaken hell-hole, and we should all absolutely go touch grass. But if you think a community effort is actually unfeasible, then I don't know what to tell you aside from "git gud".

[antinegationism]

Hey. Can you spin up a database on your local machine with 30'000'000'000 posts with perhaps ten times more tags and see how fast your search is. Please expect at least 1000 qps too.

(Context: Anon is (very stupidly) trying to poke fun at this post) Siikr currently indexes 44,325,274 posts. Due to my own implementation laziness, it does so extremely inefficiently, with each post storing a lexemetized english GIN index and language agnostic GIN index, and each post additionally containing its rawtext, json blocks, and html content representation. Furthermore, it does this with MASSIVE redundancy, because most posts are reblogs of other posts, but siikr treats them all as if they were text instances in their own right. Despite all of that inefficiency, it currently takes a grand total of 136GB of storage, running on a single virtual machine with a single virtual core and 1GB of (mostly unused) RAM.

44,325,274 posts is about 0.15% of 30,000,000,000.

If I were to instead host siikr on my old laptop (instead of the cheapest instance Digital Ocean is willing to sell me), then without making any code changes, it could easily accommodate 2% of the provided figure. If I were to modify the code to be smart about reblogs and avoid redundant representations, I'd venture to say maybe 5-15% of the figure. Which means I'd need about 10-20 laptops for siikr to host 30,000,000,000 posts. I don't know how many posts tumblr actually handles, but 30 billion seems on the low end. So probably you would need maybe an order of magnitude or two more laptops. (Note, that even this is with absolutely minimal optimizations, and I suspect that if I really wanted to I could in fact get 30 billion posts into a single laptop, first by switching to binary json or flatpack instead of raw utf8 json strings, then second by compressing post-content, and only decompressing it client side.) (Note also that this is for SEARCH. Which is a much harder problem than literally just showing people posts by users they're following, in that the latter is way more amenable to routing, cache optimizations due to the nature of viral content, decentralized storage, and compression/archiving of rarely accessed content.) On a broader point, all of the commentary on the operating costs of tumblr is friggen insane.

Like, I don't wanna be mean but, has everyone gone friggen braindead? Have we as a species forgotten that servers used to be a thing we were capable of managing prior to Amazon and Google overcharging you for them? If you force all videos, images, and audio to be externally hosted and linked to, the operating costs are stupidly cheap. Possibly cheap enough that --should there not be enough kind souls willing to leave an old laptop plugged in -- you could pay for maintenance by charging users for image and video hosting. Now, to be absolutely clear, I have no actual interest in replicating or managing a semi-decentralized community maintained version of this god forsaken hell-hole, and we should all absolutely go touch grass. But if you think a community effort is actually unfeasible, then I don't know what to tell you aside from "git gud".

[antinegationism]

The "never even notice" thing is to say that the users running the servers would not have admin privileges over them. They are just donating some some storage + bandwidth.

The bigger issue is who WOULD have admin privileges. And I think the best answer is "20 random Tumblr users are selected every three days to be demigods of Tumblr." With the the only thing to keep anyone honest being "having had your account previously smited does not disqualify anyone from becoming Tumblr demigod, and so they can absolutely abuse their power to exact revenge."

Has the tumblr community ever considered... buying tumblr? Like obviously the most recent purchase price would easily be within range of a moderately successful fundraiser, the ongoing operating costs are a mystery but basically replacing staff for dedicated volunteers would probably shave down a lot... as someone who has basically externally replaced native search for this site, would you like a crack at the actual site search? Has this been floated already? Is it a bad idea? Good but bad?

What's the most recent purchase price?

[antinegationism]

The users would never even notice it happened.

Has the tumblr community ever considered... buying tumblr? Like obviously the most recent purchase price would easily be within range of a moderately successful fundraiser, the ongoing operating costs are a mystery but basically replacing staff for dedicated volunteers would probably shave down a lot... as someone who has basically externally replaced native search for this site, would you like a crack at the actual site search? Has this been floated already? Is it a bad idea? Good but bad?

What's the most recent purchase price?

[antinegationism]

"Order-independent" here is used to mean "Exact order-independet transparency." The answer is definitely not the painter's algorithm, because that neither accounts for intersecting objects, nor gets around the sorting bottleneck. Raytracing is *an* answer, but an extremely expensive one. Other answers include super-sampling of spatial, temporal, or spatio-temporal dithered surfaces (though with significant loss of fidelity), and whatever the fuck the Sega Dreamcast was doing.

We managed to make graphics cards beat the Turing test before we've figured out how to get them to do even passable order-independent transparency. I just . . . just let that sink in. AI was literally easier than just getting shit to show up behind other shit.

[antinegationism]

I'm not so sure that the operating costs would actually be that high. If you think about it, the nature of the site is such that it's highly amenable to decentralization. You'd need to think through chat messages and some mechanism by which to avoid identity spoofing, but beyond that and the user interface, most of the functionality would just be routing information by whatever users you're following from whatever servers are hosting that information, and propagating whatever posts you make to enough servers that it can be accessed and stored. Search would prefer it if every user were guaranteed to have at least one server which always had every post they made, but even that isn't really critical.

Has the tumblr community ever considered... buying tumblr? Like obviously the most recent purchase price would easily be within range of a moderately successful fundraiser, the ongoing operating costs are a mystery but basically replacing staff for dedicated volunteers would probably shave down a lot... as someone who has basically externally replaced native search for this site, would you like a crack at the actual site search? Has this been floated already? Is it a bad idea? Good but bad?

What's the most recent purchase price?

[antinegationism]

In the unlikely event that the tumblr community manages to pull this off, I'm definitely willing to implement a proper search feature.

Has the tumblr community ever considered... buying tumblr? Like obviously the most recent purchase price would easily be within range of a moderately successful fundraiser, the ongoing operating costs are a mystery but basically replacing staff for dedicated volunteers would probably shave down a lot... as someone who has basically externally replaced native search for this site, would you like a crack at the actual site search? Has this been floated already? Is it a bad idea? Good but bad?

What's the most recent purchase price?