Is it possible to build a sustainable image and video hosting service?
The history of the web is littered with with many a dead image/video hosting service. Echos of their existence plague older forums in the form of broken links and images. It seems like they all follow the same path, starting up as the new "simple" service that just hosts images, no fuss. But then as interest grows, so do costs, and the service owners have to scramble to monetize. Generally this is done by stuffing the place full of ads until everyone leaves. Alternatively the owners are stubborn and stick to their guns, until they inevitably have to shut down due to drowning in costs. When they do shut down, millions of assets are lost and the graveyard of broken images across the web grows some more.
https://gfycat.com/ is the latest notable victim of this.
With all the recent social media turmoil, there as been lots of exploration of alternative sites, and all of them have to overcome the problem of hosting media in one way or another.
Tildes obviously does this by avoiding it entirely which, while a very effective solution, is just handballing the problem elsewhere. Users will still want to post images and videos but they will just have to find alternative hosts. Over time those hosts will die and Tildes posts will be filled with dead links.
Mastodon has similar problems,the biggest cost of hosting a mastodon instance is the storage and bandwidth required to facilitate media posts. And there's a real danger of an instance incurring high costs if a particular post becomes popular and is hotlinked on a big centralised social media site.
It seems like a really tricky problem to solve, something peer-to-peer could sort of solve the costs created by traffic peaks but has problems when there is many small files viewed by few individuals each.
Are there any other solutions out there? Web3, IPFS? Or is it just not that much of a problem, do we accept that media on the web is ephemeral and will be lost after a while?
It is, if you build in lifetimes. Everything has a lifetime, and ignoring that leads to trouble.
Text is less sensitive to this because it compresses extremely well, but images and especially videos will quickly run up the limits of "paper sustainability": where something is only sustainable because it hasn't reached a certain level of activity yet.
And the thing that comes to mind here is, funnily enough … 4chan. It ran pretty damn sustainably for such a notorious image board, and that's to do with how lifetimes were built into anything going into the site.
The thing with image and video hosting is that by definition, the content is mostly worthless. If the platform is free, then you're just providing a free service in exchange for zero value. So for it to be sustainable, you need a well-oiled machine behind it that funds it: Investors, user value extraction, re-selling of data, etc. If that disappears, so does the money, so does the service.
The other way is to limit the intake, make the content worthwhile (edit: @staross mentions that, didn't see in time!). Nebula is a great example: At it's core, it's the "same" content as on YouTube (it's literally youtubers, people who have made youtube their career, who are on it). But the content is limited and curated, which makes the content worth more, thus users are willing to pay for it and now you have something sustainable. Bandwidth and hosting costs are bounded, and never exceed what the users pay. (Although lifetime subscriptions, which Nebula did, are a great way to break that sustainability - but they did it in a pretty sane way so, whatever).
Thanks for that link, it's a good read. I think because I come from a family of librarians and archivists I have a tendency to consider most information having a very long lifetime, and so I find it hard to accept that content created on the web can just disappear, especially when there is no way of quantifying what value is lost.
In regards to your last point, I think it is important to consider that the worth of content is not objective, and also not constant. What seems worthless today might be vital in the future. Marion Stokes is a good example of this. At the time she was recording TV it was probably considered a waste, but years later it became a fantastic, and unique, resource. YouTube videos that are low quality and have <100 views today could provide amazing insight in 100 years time, it would be a pity if they were lost.
I’m an archivist and absolutely agree with you. However, as an archivist, it’s especially important to recognise sustainability as critical, and the ability to work within the bounds of a capitalist framework.
Uncurated and curated content both provide different types of value, but either way the value needs to be extracted and offered for the sustainable aspect to hold its own. Uncurated content for example can provide an amazing first layer of data for AI labelling.
But for example, I have several terabytes of (video game) data I have tried to spread around. That data is worthless until someone does something with it. I do not have the time to do something with it and nobody has taken me up on it yet. I’m at this point the last person in the world who has such a complete dataset of this particular type. And yet it only exists and is accessible because I choose not to delete it. Everywhere else it’s been made accessible is essentially just … archaeology territory already: “if someone finds it and recognises it for what it is”.
So… yeah. Useful, but useless; and the useless part means nobody will care once I’m gone, unless it finds a proper home. Lifetime of that data is currently equal to my hard drive’s.
I already have too many projects in my backlog, but it's also not clear to me what this data is, also from your comment below. You describe it as being about video games and that it looks like garbage and that it's huge. Like, eh okay, would you want that dataset if you were casually reading this thread? I have no idea what this even is, assets? Gameplay recordings? Source code? Of what type(s) of games? Before someone can think of a use for the data, you'll have to describe it
If you just want it to be archived, I think archive.org would be my go-to place for it
Which, on the topic of "everything has an expiry date", I wonder how that's going to be sustainable, but so far they've managed for longer than most image hosters were around, and they store more data than afaik even the biggest image hosters (to be fair: they also have lower bandwidth costs because it's primarily an archive and not meant for embedding)
My goal wasn't to specifically share this dataset with anyone, here. I'm better placed than most people on earth to know who might want it, and I've resigned myself that it'll probably die with me unless every possible star aligns.
I’d be very interested in that data, and depending on what it is I may know someone interested in it!
Blizzard uses a protocol called CASC, it's kind of git-inspired. I wrote a client for it and until 2020 ish, I was gathering all the data. They're a mix of encrypted and plain data blobs.
The data is... utterly useless if you don't know how to read it. It's /work/. Because there's also no client built to actually browse this data. There's no client built to find the missing pieces, to verify the data and ward it against corruption, or to redistribute it as a server. I was working on an IPFS-based solution but it never went anywhere and now, I'm busy.
The data is invaluable, but without this work, it's also completely useless. It just looks like random garbage.
If you really want it, I'm happy to share a link. There's IIRC ~1.5-2 terabytes of data; incl. millions of small files. But I guarantee you that you won't be able to do much with it, just like you wouldn't be able to do much if someone showed up with ~500 barrels of crude oil at your doorstep.
What are you planning on doing with it? Or, what is the value that it has?
If someone were trying to reconstruct a specific patch version for a specific Blizzard title, would be my guess.
I could see it potentially useful to private gameserver modders trying to quash bugs in clients.
I studied history in college, and it's a regular pain point when you are studying ancient history how much information we've lost. Look at Ancient Rome, for example. In some cases, we have books that make offhand references to popular, comprehensive histories written by other people. But we don't have the books that are being referred to. (For example: 107 of the 142 books of Ab Urbe Condita, a history of Rome by Livy (59 BC – AD 17), are lost.) They're gone – presumably forever.
Historians of the 21st century are going to have a similar problem, particularly if they are trying to study the everyday lives of normal people. Yeah, our most popular books, movies, games, and TV shows will likely be preserved indefinitely (though our storage media can sometimes be problematic for that). But what about YouTube videos? What if someone in 2242 wants to write their dissertation on meme culture of the 2010s. Will there be anything left for them to work on?
If Imgur ever went down, there are so many Reddit threads that would be virtually incomprehensible.
It's possible but you have to get very lucky. If you build your site with revenue generation built in from the start and can afford to scale it as you go so you're never dependent on external investment, it's do-able if your business is built around running sustainably and not chasing profit at every possible opportunity. Whether that's ads or subscriptions or whatever - that's where you have to get lucky, hitting on a revenue stream that works from day one and scales at the same rate as your costs.
The problem comes when you suddenly need half a million currency to buy a stack of storage/processing/transfer/whatever and you go to an investor to get it. Then they want returns and will push for you to do the thing that makes them the most profit. Nobody is going to hand over investment cash for a business plan which says "grow slowly and carefully and only make enough money to pay costs and reimburse staff reasonably", they're all looking for "rapid growth over 2-3 years then IPO the fuck out of here and cash bonuses for all VCs"
Nebula managed it. It was founded and run by experienced video content creators. So far they haven't needed VC capital, but at one point they had a partnership with a more well-established company (CuriosityStream).
Wendover Productions recently made a video about this: https://youtu.be/Alqt6RCEWdM
Not that I can think of. All these services suffer from the same issue in the end: They are expected to hold an infinitely growing amount of data as they expand their operations to support holding more data. The only solution would be to completely remove any profit incentives from this and run media hosting as a public service, effectively printing money to pay for the infrastructure to handle this ever-increasing burden.
I don't mean this as a politicized or ideological talking point, but simply as a practical point. We would need a massive international network of public commons to hold all of this data in a growing trust to ensure any references to it are preserved for the future.
I've been thinking about this lately. For a community-centric site where the site is not owned by a for-profit business, I can imagine a hybrid solution where recent data is stored and hosted normally. Say, up to 90 days.
Afterward, to avoid those infinitely growing costs, the community would take over hosting, with the assumption that images are accessed less frequently as they age. (Making exceptions as needed.)
BitTorrent or IPFS would handle this second storage layer. The main server would request it and still serve it, like a proxy, so users wouldn't have to connect to random IP addresses to get older files. There would be some expectation that the original uploader would stick around and provide this "permanent" copy of the image; and there will surely be some community members who donate their storage and bandwidth too.
I'd opt for tiered instead, you could go as far as going for High Res on SSD, SAS, then recode to highly compressed for aged material and finally out to long term on LTO and leave a piss poor proxy as a placeholder. If people are willing to wait, LTO robotic libraries are there and can retrieve items but it adds a delay of minutes so stuff could be watched back in higher res at a later date, you'd just need to tell people if they wanted to see higher res they'd have to click the link and wait, or check back.
All of this is still expensive though.
With the new formats dedupe is really difficult, but at least identical pics and videos could be which would give another element of compression.
If you don't mind the data loss you could put the original to the uploader but most people upload and turn off their machines. Stuff will be offline a lot.
In the Netherlands, we have a national version of this: https://www.beeldengeluid.nl/en/about
They have their own PeerTube instance, where you can watch some of their collection: https://peertube.beeldengeluid.nl/
(They collaborate with other public services, like libraries, including the music library)
That handles data retention, but not really the issue where a service dies and takes its links with it. I guess my idea would be that this repository would be used as a primary facility for hosting gifs and stuff to ensure continuity of links posted to forums and whatnot.
That Dutch solution is pretty great, but doesn't exactly solve he problem.
The US has the library of Congress which does record and store media but it's far from comprehensive and access is subpar.
I feel like this is solvable. I’ve long believed that content on social media should have a default expiration date that needs affirmative effort by the creator to maintain or preserve instead of defaulting to retaining it indefinitely until the host runs out of money and pulls the plug.
You could have content simply expire after 30 days and make that a clear expectation up front. You’ll lose market share to all the free services who don’t do that but they’re all torching money anyway. Someone has to be left standing.
Something about the digital world makes everyone o to hoarders who can’t just let things lapse or be forgotten. This applies to text too IMO, largely for privacy/right to be forgotten reasons.
Hosting images and video would save a lot of money if it was efficient at the get go.
Max resolution set, avif or webp, av1 or h265, and a decent encode. Browsers are the problem, they don't keep up. Actually, it's the W3C who set standards don't keep up. Everyone wants backward compatibility but the pace of tech is evolving beyond setting a standard every few years.
There is the main issue though, standards and compatibility. Those are what cost. If you upload to YouTube it makes multiple versions of the video to host to a plethora of devices. That is where the real cost lies; the compute, the storage and then the pipes to feed it to everyone.
A 10Gb cable in the UK which would work for the most part is only £10k a year. The rest is dependent on how hit up the service is. Call it £50k capex to build a server farm with a petabyte of storage. Now you have to pay Devs, sysadmins, etc a salary on top.
Everyone blocks ads, they are not a good way to get paid anymore. What are you left with? Subscription?
Or p2p options. BitTorrent worked, even in the early aughts, and works even better now that more individual nodes have fatter pipes. I think it's a huge tragedy that centralized media distribution channels somehow became dominant in the Web 2.0 era. There's been a bunch of interest in somewhat decentralized social media lately with Mastodon and Lemmy etc., but a decentralized, hash-addressable, general media sharing service makes much more sense to me. Maybe for power-users, paying for priority access to peers that would act more like CDNs would be nice, too. There's lots of room for innovation there, I think.
There are issues with IPFS though. If only one machine is holding the data and goes offline, it's offline. Now you need to have replication which is duped storage. This is the point where you start thinking that you may as well as have static or geo-replicated data and call it a day if you're a business, you have to give over the guarantees.
I'm not saying it's a bad idea. IPFS would work if it guaranteed an online copy all of the time but it cannot, unless you also have one node which is the master with a copy that is owned by the business. As soon as you do that though you just took most of the costs again.
From a business stand point, you will have to punt the cash for the compute, distribution and storage. The main concern is balancing the cost of the service Vs money coming in to at least break even. No one wants to break even though (apart from @deimos it seems) as businesses want to turn a profit.
But are there any successful applications or services that have been built on IPFS?
I do think its possible, but would need more research and thought put into it for it to be feasible for video sharing. My initial thought is how we can stream chunked data, that assumes it will be loaded in order. Usually when I torrent I have to wait for the whole file to be downloaded for it to be playable. Maybe a fediverse-type version would be better, where video/image host instances can connect to each other as needed. I do think instances would have to charge a subscription, though. Either that or hosting a video host would have to become a big trend to account for all the data storage needed.
Streaming from torrents was already done a while back with Popcorn Time and some others. Granted, it only really works when the swarm is large enough. When you don’t have enough peers with the chunks you need, it falls apart (or you have to sacrifice bitrate etc.).
Those storage and compute costs are of course huge when one provider decides to handle all of them. Back when YouTube was new and video content was not common (or well-implemented) on the web, it made sense for one service to handle all of that work.
Now that anybody can throw video files on a physical or cloud server and host them from their personal website with a few dozen lines of JS, I don't think it makes sense to centralize ownership under one roof any more. I'm really optimistic that federated video could become a huge thing in the next few years. With one decent JS library, you could handle transcoding, subtitles, playback speed, and more all in the frontend! Just serve up the 4k files in a decently compressed format like h265, keep server costs minimal, and let the clients handle the rest.
YouTube clearly can't create a sustainable business model that doesn't rely on exploitative ads, poor content owner compensation, and needlessly expensive subscriptions. But I'm sure a lot of content creators could band together into little co-ops, maybe assisted by something like Patreon, to run a federated video hosting instance. Discovery is a whole separate can of worms, but I suspect social media can handle that in a healthier way than today's YouTube algorithm which... clearly has perverse incentives.
This is pretty similar to what Nebula does already btw.
The challenge is discovery. YouTube has a powerful recommendation algorithm that directs viewers to your content. Self hosting means you need some other avenue to attract viewers. One upon a time that might have been possible but nowadays so much online activity is centralized on platforms designed to keep you trapped inside then there’s no way to organically get eyeballs on your content.
Did anyone else know that Gfycat was owned by Snap Inc (the company behind Snapchat)? There is a shocking lack of info about this, looks like it acquired after Meta's buyout of Giphy in 2020 (1). If they weren't willing to subsidise it...
Just spitballing here. Imagine a video (or image) hosting service where each piece of content has a "grace period." Maybe it's a week or a month that the content will live without any intervention. Each piece of content also gets its own donation button. As donations come in, they increase a life counter on that piece of content that transparently displays how long the content will live on the platform (after the grace period). Once donations stall and any remaining time expires, the content is deleted.
Is this anything?
Snapchat-esque for pay.
It's unique, I like it. I can imagine the famous Putin and China tank images living forever with donations.
Idea wise it doesn't seem too bad, although you'd need to find a way to make those donations as effortless as possible to encourage people to chip in, which I assume would be the actually difficult part if this were to be implemented.
What I'd also throw in is the ability to donate to the entire backlog of an uploader and have that "spread out" across their uploads, which would solve the "helpful person posting screenshots on support forum" use case.
Yeah. That is easily sustainable. You'd reach a peak size that just turns over.
It's how companies like WeTransfer operate. 2GB free upload for a week holding.
If you want anything out of that, it costs.
Applying the same strategy would work. It semi takes care of DMCA takedown and GDPR requests too.
Catboe.moe is great. I just became a patron.
There are some great ideas in this thread, but a lot of them would only curtail storage costs, when it's bandwidth and (particularly for video) compute that are often the real issue.
Getting video into a usable format - a cost that comes up front, even if only one or two people ever watch it - can easily cost more than storing it for 6-12 months does. That means you've got one very significant expense that scales with upload volume, regardless of playback volume.
For content that does become popular, the data transfer cost rapidly overtakes the storage and transcode cost. So now you've got a second significant expense that scales with playback volume, regardless of upload volume.
Significantly limiting either of those two things is basically the same as limiting usage of your platform - which is fine, depending on your goals, but I'm not sure it would meet the definition of sustainable if the platform simply can't grow past a certain amount of usage, which may be met simply by linking something on a website that's too popular?
Distributing the inevitable costs seems the only plausible option, in one form or another. I like the P2P idea, because it directly alleviates the actual resource contention - people chip in storage, compute, and bandwidth rather than the cash costs of them. Since a lot of people have those resources sitting idle, it seems like a win win. If that's not viable, maybe for content moderation reasons, then I can see an uploader pays model being just about workable, especially if it were done more or less at cost - but even for a few cents, the friction of signing up would cut usage a lot. Maybe that in itself is a good thing, I don't know? A hybrid approach where large websites bulk purchase usage credits for their users could smooth that out, but then we're just building S3 with extra steps. Of the options out there, the idea of downloading a system tray client and getting access to everything strikes me as a lot more egalitarian.
A lot of people are talking about infrastracture, but moderation of multimedia platforms is also an issue. For instance, if you're in the US, in order to qualify for safe harbor status, you must have a way for copyright holders to make DMCA claims and you must act on those. If it's just you and a server, you can easily become flooded by DMCA claims, and either have to spend all day processing forms, or just give up and eventually the site will be taken down.
Youtube has the scale to build algorithms for this. You probably don't. Not to mention people don't seem to like said algorithm very much.
Another is CSAM. 4chan is absolutely going to upload child porn if nothing else just to screw you over. You are also expected actively monitor and take these down in you're in the US. How are you going to do that? Are you going to do that manually? Did you budget in therapy? There'll also be child porn with gore! 4chan will find it.
All of this gets magnified, in terms of compliance, in any kind of p2p system as well. Many peers won't be very happy unwillingly hosting illegal material.
Maybe don’t be in the us, and the do reasonable effort to remove child porn, wares, and the like, and you’re probably golden until you’re big enough to have the revenue to do it more accurate?
The problem is you need an absolutely massive number of labor hours to tread through all the uploads. Imagine if Reddit had to pay every moderator. Where do you get the funding after a certain point? Donations simply don't scale up to funding an image and video host for the entire internet.
Complying with international laws gets very expensive very fast, and you'll absolutely get taken down or restricted if you don't have strict limits. Look at PornHub for an example of what happens when the Christian zeitgeist finds a new way to attack a service. Content doesn't have to be obviously illegal to open you up to litigation.
I feel like we're going to keep seeing services like these shut down as long as we continue to believe that everything on the internet should be free (as in beer). I've found my thinking shifting over the years to want to pay for the services I value in order to contribute to their continuation. I am hoping that whatever new era of the internet we're entering will involve less user-as-the-product services like Facebook, and more user-as-the-customer services (like paid email providers). Not that I just can't wait to pay for things, but I think it is a much more honest business transaction to just pay money for something rather than get it for "free" but then suffer the relentless invasive monetization.
I don't often need to post images on social media, but in the past I've used Imgur to post to Reddit, obviously for free. I can imagine a world where Imgur charges a fee to use the service on an as-needed basis. I also remember the old days of the web where your ISP would give you some small amount of web hosting as part of your plan. I can easily imagine that type of service coming back, as an add-on to something people are already paying for (Apple iCloud, Google One).
I agree with you! I'd rather pay for a service than have myself be the product (have my data harvested and sold to advertisers)
Honestly, you are going to need ads or subscription revenue. It costs a lot once you get into TB range. Would also be good to have AI image matching to delete duplicates on the backend.
This is why I thought it was reddit's beginning of the end when they announced image upload functionality. It had been incredibly expensive to host them all for free/donations/ads. The only image hosting that had been stable to use on forums (it was that era) was self hosting or paying for image hosting.
But then, it seems reddit hasn't started deleting old materials yet. I reflected last year to my gf that disk space and bandwidth costs must have deceased enough for this now to be possible.
But it still requires a proper business that makes money to pay for this. With gfycat being mainly embedded use, at least in my experience, I don't see what their business model was in the first place
AFAIK one reason they did that was because Imgur had threatened to block them, as a sizable chunk of imgurs traffic came directly from Reddit to hotlinked images.
This entirely relies on what you find “sustainable”. Should it remain free, as most services currently are?Should those services remain reliable and with the same uptimes? Should those services continue to provide near unlimited storage space? Should they continue to be moderated to remove any illegal contend? Should they continue to provide amazing features? Should the data remain available forever? There’s so much these services provide other than just content storage.
All these things cost an unimaginable amount of money. Just hosting this reliably globally sounds like an Impossible task to me, which will already cost an incredible amount of money. And then there’s all the other stuff listed above. Users are relying on all those features and will not like to downgrade, so as of right now you would be stuck with cost. The way the services are currently solving this is through venture capital, selling your data and ads. And they are still operating at great losses. Most will be aiming for market domination so they can start fully monetizing your presence on the platform.
In the car community there’s always a variation of the following being mentioned: a car can be cheap, reliable and fast, but it can only be two of those at the same time. I think the same is applicable here. If you want to have forever storage and have it be a reliable host, it will be expensive. If you want it to be free, you’ll have to give up on one of the other two. And that’s just three of the things mentioned in the first paragraph. You’ll have to repeat this process for all the things and decide how important it is to keep them.
The short answer would be, it is way too costly to provide this service for free. In the way (and at the speed) we currently are adding content to the web, it is impossible to create a service that will archive everything forever and be free at the same time. And do we care enough, to start putting extra resources into it to achieve it?
We substitute images using ASCII art.
On a more serious note, this is only viable for smaller sites, but the host should maybe back every media up, and find another way to at least keep the media up whenever the image or video hosting sites die.
In all seriousness, perhaps ASCII isn't far off from the actual solution, and the "solution" is a radical downsizing of memory usage per image - like, memes are in theory often only a handful of bytes, given they're just an existing image with text overlayed. Also, obviously, vector art exists, it just needs a good interface for "uploading" "images". There's surely some threshold where running an image hosting service is viable - like perhaps a 5KiB cap would make it viable? (although, "viable" under which particular business plan?)
I put "solution" in quotes because it solves the problem by changing the problem, and obviously it wouldn't function as a general-purpose image-provider.
Haha, I like the idea of ASCII art. Storing a local backup seems to be how mastodon works, if you look at a post originating from instance A on instance B, you'll be served a copy of the media that is cached on instance B.
I wonder if it would be possible to distribute media storage across multiple instances. ie. if one instance federates with many others, when a media post is loaded on those other instances they take on the responsibility of storing and serving a chunk of that media. There would need to be some sort of integrity and redundancy system but it would mean that each time a piece of media is requested the load of serving it could be spread across many instances.
There is a way to build a sustainable service provided we focus on "live streaming" instead of storing/hosting them directly. As you said, the storage costs will be monstrous today as even a few videos in hi-res will soon run into terabytes.
On the other hand, if the server can simply stream out whatever you're feeding from your desktop or mobile camera and not store them, it'll be both scalable and sustainable. It is quite trivial to build such a streaming script in a language like PHP and host it on a low-cost or even free-tier hosting service. It will be up to your fans to store and archive them if they want it. Simple idea but win-win for everyone.
That's just Twitch. And they too are suffering from massive financial issues, which is forcing them into excessive monetization that is pissing off the users. I don't believe their model is sustainable. And I doubt just deleting the VODs would change that.
This isn't a storage issue as much as you think. It still costs ridiculous amount of money to push enormous amounts of data to millions of people, all the time, 24/7.
Also, there's the issue with content creation that everyone is forgetting. Video doesn't just magically appear out of nowhere. Most content (especially quality content) comes from people who have turned it into their full-time job. They demand to (and frankly should) be paid for their work.
The cost of that cool YT video you just watched isn't just the few hundred MB of storage it takes, that's only a small part of it.
Twitch is suffering because it's centralized. What I'm proposing is individual creators themselves managing their IT with a hosting service. Pretty sure there must be an open source alternative to twitch, if not it's trivial to write a PHP script as I said.
For monetization, everyone have their own preferences. They can put up patreon links or they can team up with an advertiser and monetize the affiliate income.
https://owncast.online/
https://github.com/owncast/owncast
That's a really cool piece of software, yep! Also surprised at the use of golang, most mainstream FOSS web apps are usually written in php/python/rails. Didn't knew there are so many popular golang apps.
I'm all for self hosting, but a trivial script isn't going to be usable, secure, stable, or fast. The library that @ourari linked looks like a great option, and that represents ~40,000 lines of code from over 100 contributors across a three year period, for example.
Even assuming each content creator does manage to find and set up a software package that works for them, data transfer is going to pretty rapidly outpace storage costs as soon as they have more than a handful of viewers, which renders the "stream, don't store" part of the idea kind of moot.
The key thing self hosting would do is push each individual creator to cover their own baseline costs: right now, Twitch is losing money on transcode and bandwidth for tens of thousands of near-empty streams in order to be the platform that the big, profitable streamers gravitate to. Similarly, YouTube onboards terabytes of content every minute that'll never get watched just so they can keep their place as the undisputed one stop shop for user generated VoD. Push those costs back to the uploader, as they would be in a self hosted world, and profit margins would rocket - but usage would also fall off a cliff, and that's not something they consider acceptable.
What about deploying IPFS as a "utility" service, the same way that we do DNS? It's in Internet providers' best interests to minimize the amount of backhaul on their networks; might there be a point where the savings justify the cost to deploy it?
Do you mean GDPR articles 17 and 19 (Right to erasure)?
I think all this does is lay bare the weakness in the current regulation. If one looks at IPFS as a big distributed cache, with the node on which data is pinned as the "data controller", then unpinning and purging the cached data is erasure.
If you imagine a similar rig, in which, say, an individual, CDN, or archiving site sets up a caching HTTP proxy to cut down on bandwidth costs - what happens when the origin site takes down the content?
EDIT: good discussion here: https://discuss.ipfs.tech/t/ipfs-and-gdpr-cpra-compliance/13978
Maybe severely limiting what gets saved would help. I imagine that a large portion of these media are garbage, reposts, etc. Not too sure how that would be done technically though.
I think you would need some kind of paid for subscription to give you the ability to use the main host for media storage and an allocated account with good tools to help you keep on top of your storage usage.
I'd point to https://cubeupload.com/ as a good free image hosting without bloat that is supported by donations.
I think that's the first time I "advertised" this website and I'm not an IT person. I wouldn't be scared of leading anyone to a bad path on ~comp anyway, you guys clearly know more about that stuff than me ;)
I only used cubeupload a handful of times, I like that it allow direct links to work as expected on mobile -- imgur makes me load their entire website.
Funny you mention this, I just got a bunch of old photos off my photobucket account. They’ve been sitting on photobucket for over 15 years. Photobucket has now moved to a subscription based service. I did the 7 day free trial.