this post was submitted on 02 Nov 2023
81 points (89.3% liked)
Android
714 readers
3 users here now
Android news for android developers. Everything that happens in android world.
For Android development specific topics please see /c/android_dev
The Android robot is reproduced or modified from work created and shared by Google and used according to terms described in the Creative Commons 3.0 Attribution License
founded 1 year ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
You underestimate the cost and complexity of video streaming.
I can assure you, I do not.
I've heard that yt handles around 3PB of new uploads daily. A 10TB drive is optimistically ~250$, so if you want to seriously compete with youtube, without taking into account data redundancy and, you know, actual servers and traffic, you're looking at a bare minimum of 75000$ of new hardware each day. No one can afford to burn that much money other than google.
I've made most of my arguments in other replies, but here I'll just say this: Most of what is on YouTube is worthless garbage.
They've developed a business model around being the garbage collector and storage provider for home movies, dipshits that want to be viral by pranking people, your cousin's drum solo during their recital, every awful cover of all of your favorite songs, and some guy's unhinged political rant (recorded in his den at 2am)
I think maybe it would be ok if they posted their stuff to some federated provider that charged a few bucks per gigabyte. They sure wouldn't lose money on bandwidth; the videos wont get viewed more than once. If that.
As for actual creators? They'll be able to self host, or band together and make mini-services funded by like-minded fans (and probably some sponsors, because capitalism), and everyone will be able to access everything on an interconnected... what's the word? Oh yeah, "Internet". You know, the thing we had before 12 companies took over everything.
Those creator services exist (e.g. nebula) and are great, but they usually cost money, because video hosting is apparently too expensive to just run on donations, and competing with google on advertising is even more of an uphill battle
I currently support Nebula. That was easy money to spend, unlike the prospect of giving YouTube anything.
Also - since they actually curate their content - there's less of it, and higher quality. Kinda speaks to some of both our points. If they had a policy of "free to all, after a while" (like a lot of patreon people do), they might well have attempted some kind of distributed hosting. Hard to say for sure, but a guy can dream.
Federated video streaming may not have all the questions answered right now, but people are already attempting it. I think the right optimizations, the right content, and the right audience will push it really far. And maybe it won't be "YouTube quality" for a while (or ever), but who needs 4k60 Minecraft/Fortnight lets-plays anyway?
There are already lots of places to post low value video. Basically, every social media site out there let's you post the home movie you want your friends to see and don't care if any stranger ever sees it.
I have a YouTube channel. It was created as part of an experiment that failed. I think I might have a total of 4 videos posted there. If it comes right down to it, the traffic I expect for my personal projects means I could just post any videos I create directly on my website.
And that personal website is where we need to get back to. I wish all the fancy programmers at Meta and elsewhere would just put their energies into creating the tools that let people put content on their own site as easily as on Facebook. Add some semi-structured data that can be leveraged by displaying the results of pre-built custom web searches and you should be able to approximate the experience of Facebook.
That's one of the use-cases I've had in mind. You have your site, you have a video. If you posted your video to your site though PeerTube (or really any implementation of ActivityPub), both your direct site viewers AND anyone searching PeerTube (et al) would see it.
If lots of people did that, you have a basis for a version of distributed YouTube. Small creators' (or people just messing around) videos might load slower or only have lower bandwidth options (resolution), but larger, focused creators (CorridorCrew, Kurzgesagt, SciShow, etc) would have more options through viewer support and collaborations.
Yes, that fits my own mental model of the online world I'd like to see.
In that case, how can a federated youtube handle the significant traffic and storage requirements?
Literally the same way any other federated service works. Here's a list of over a thousand instances that seem to have figured it out;
https://instances.joinpeertube.org/instances
What has the amount of instances to do with that? Everybody can open one, that means nothing. And what means "figured it out"? It don't see massive traffic there.
The largest instance has 20k users (and signups closed). youtube has 2.7 billion active users. now, this instances does not give any information about the used hardware, the second biggest (AntTube) does though.
Looks like its running on rather low-spec Core i7 and 8 GB RAM. The site does however say nothing about how much the hardware is utilized and also not how much bandwidth is used per month. So what are the cost here?
And more importantly, how does it scale up? How much would PeerTube need to serve 100k user, 1 Million? 100 Million? Is it optimized for that?
You massively underestimate the development effort and the know & how that is in Youtube.
And don't even get me started when it comes to making money with it. Some companies will not just advertise on some random privately run sites. Apart from the fact that they are legally probably not even allowed. There must be a real company behind it with which you can conclude a contract and which can be held liable if something goes wrong.
So there only need to be 135,000 peertube instances to hit YouTube scale? That doesn't sound like a big deal. It's not like they will all magically appear overnight, but neither did YouTube's scale.
Everything in the fullness of time. One person sets up an instance for fun. Another sets one up for ideological reasons. Then someone else sees setting up an instance as a practical matter, because their family doesn't need to put up with ads just to share home movies. That creates enough of an ecosystem that some people set up as a way to extend their reach or hedge their bets. Someone else realizes that their real income is from through paid subscriptions, so it's cheaper to run an instance than to pay Google their take. And so on.
Maybe it fails, but accepting failure before making an attempt is one of the most insidious flaws in human nature. It should be battled at every opportunity.
Good, I'm glad advertising will suffer. Even on YouTube, creators don't make shit from platform advertising. Their best shot is either premium user watch time, or direct sponsors (heavy emphasis on the latter).
The number of instances is relevant because it illustrates that this is already a maturing project, not a proof of concept. It's not even the only federated video ~~platform~~ project.
As for the how; I'm not going to sit here and develop a roadmap for you just because you think I'm being reductive. Yes, it's complicated, but it's also not magic. Adoption breeds innovation (more eyes, more devs, more complaints), and necessity breeds refinement. Want to reduce file sizes and make the most of bandwidth? Develop/refine codecs (see h.264 vs AV1). Not satisfied with the speed of encoding, or want to squeeze more frames out faster with existing hardware? Refine the encoder. Don't want latency for distant users? Make sure the app knows how to find closer peers. These are just natural refinements made over time, same thing YouTube did.
Which brings us to monetization; First, maybe advertizers /should/ be a little picker, and more careful. Hell, there are news stories alleging that Google itself isn't being totally honest about how ads are being shown, or to how many people. And also, fuck advertising. I'll pay for a well run federated instance (or run my own) before I give YouTube one cent after all the shitty things they've done (their terrible handling of DMCA comes to mind).
Being hard doesn't mean it's not possible, or not worth it. Even if it's slower, or lower resolution, or "only" 24/30fps, it's well worth it to get out from under these fucking monolith providers we've chained ourselves to.
Yet they still can’t stop adblockers. Despite what this article says I have no problems with ublock. All my friends ublock still works.
Youtube coders know more about “Cracking the Coding Interview” than they do about making good software. Just a bunch of egotistical greedy little pigs working for their perceived reputation and money.
Well said. I've lost track of how many times I've seen shitty code go into production with the justification "we'll just throw more hardware at it". There are a MASSIVE amount of resources that could be reclaimed if we went on a diet and stopped relying on bigger and faster servers.