this post was submitted on 18 Jan 2025
235 points (89.6% liked)
Technology
60566 readers
3594 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related content.
- Be excellent to each another!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, to ask if your bot can be added please contact us.
- Check for duplicates before posting, duplicates may be removed
Approved Bots
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Video will not work for AT whatsoever. Text and images, fine, but I'm pretty sure leveraging edge delivery of video is just not going to work out well for users. I think they'll need a centralized host for that portion, or some fancy ways to offload bandwidth otherwise to prevent constant hammering of popular videos.
https://en.wikipedia.org/wiki/PeerTube
Yes, and it's incredibly slow and wouldn't scale to millions of users. If one user is high bandwidth, and another low, you'd have uneven distribution of traffic for a newly connecting user, meaning the entirety of whatever you're about to watch won't be completed in time for a good user interaction flow. The issue isn't whether it's technically possible or not, but if it's functional enough for similar traffic as TikTok.
The other issue with torrenting is that a lot of users may incur data charges if the service were to be constantly seeding other users on limited data plans or with data or speed caps in general. It's just not the right tool for the job.
It will scale just fine, so long as the ratio of instances:users is similar.
The current ratio of consumers:creators on youtube is 41:1, by my research. A single server of sufficient power could easily serve thousands of users.
That's some shitty research you've done then.
1000 users streaming something that's 5mbps would be 5gbps.
5gbps isn't common for consumers... and costs a lot in a datacenter (about 4k/month on the cheaper end).
Buddy, do you not know how periods work? That's 2 different sentences you've mashed together and pretended they were one.
Secondly, I didn't say simultaneously.
I do know how sentences work. I also know that paragraphs and posts sound be related to each other. Your sentences are not completely divorced from each other.
The point was that you're claiming to do research on something just to turn around and say something that WILDLY wrong. This discredits any amount of research you would have done.
Doesn't matter if you say simultaneously or not. You said THOUSANDS... I showed you just 1000. And this was ONLY looking at bandwidth. Not actual server costs.
I claimed to do research on something very specific. If you have evidence to the contrary, please feel free to prove me wrong instead of just intentionally misrepresenting my statement.
...of course it does? A thousand simultaneous streams is not going to have the same load as a dozen...
So now you've backed down from "thousands of users" to a dozen?
If you have THOUSANDS OF USERS (your words)... you should probably at least plan for 1000 concurrents, probably more (remember you have to plan for peaks, not average).
You seem to be missing this repeatedly... I'm not sure how else to present it to you. You made the claim that a decent singular server should be able to host THOUSANDS (with an S... so multiple thousands.) I'm showing you that even if it's just 1000 concurrents, you're paying a heavy cost JUST for bandwidth... forget the server. You're over your head if you think a single server is doing this shit.
I run a plex instance, I have 8gbps internet to my house. I could host probably 80-100 simultaneous streams on that bandwidth of raw blurays. My servers could not handle that load simultaneously (and they're hooked up as 40gbps internally). If bandwidth is the easy side of this equation (it is)... and your assertions are already failing... Then you're just plain wrong.
You continue to engage in bad faith strawman arguments and try to misrepresent my statements, even after I've already clarified them, so I have no interest in continuing this discussion.
I haven't bad faithed anything... You clearly don't understand how hosting a service works, in the case of "thousands of users" 1000 active is a hard lowball.
You've clarified nothing. You constantly moved goalposts and pushed random "facts" like those statements changed anything about the original premise you presented and my response to it. You started with "THOUSANDS OF USERS". Then backed down to 12 two posts later.
THOUSANDS OF USERS -> Okay so let's take a case of 1000 active users...
UH UH, I didn't say simultaneous!!! -> Good thing I only took a case of 1000 active users then... BTW we're not even looking at server costs for processing, just raw bandwidth.
Uh Uh, What about 12 users!!!!!!!! -> (we are here).
You need help dude. Nobody is coming after you. And nobody misrepresented you. You're just completely out of your depth, which is okay. But don't act like somebody is misrepresenting you, the world can read your responses.
That's not how this would have to work though. Even with dedicated seeding instances, the instantiation of a session for a torrent is LONG. Like 5s+ long. A request and response from a CDN is in the milliseconds. Users wouldn't use a system that takes 5s just for the initial request for a single video, plus the additional time to sort for segments and recombining before it plays. Even in a fast-ish scenario, that's like 10s alone.
Imagine waiting 10s for a stupid internet video to even start playing to watch some kid dance with a rubber chicken in their pants.
That's weird because it works instantly for me.
Could implement torrents, which I believe is how Peertube handles it.
You apparently do not internet. Anything with a wait time is not an app platform people will use.
and it’s not outlandish to suggest that length decreases every year.
109ms response time was the goal circa 2009. Now, with the advent of Kubernetes, response time as are in seconds. Proxies on top of proxies cost request time. Users will abandon.
it's called a CDN
CDN won't scale to millions of users all uploading videos on a decentralized system. Article is specifically talking about AT Protocol which doesn't account for video. Making a global CDN distribution of videos from decentralized sources is whole other ball of wax.
Bluesky has videos though?
Bluesky has videos linked to a static host by a different site or provider. They are not themselves the host of those videos. Different scenario.
So why does it for for AP with Loops? What’s the fundamental difference between, isn’t the Fediverse the more decentralised system?
Because Activity Pub is not a data salad. A video lives in a specific place.
I don’t think it has been proven that Loops is viable at scale yet.
Can you reform this question?
Loops uses ActivityPub and feeds video in a TikTok like manner.
What's the difference that makes it not achievable for AT but okay for AP?
Long story short Activity Pub only pulls the content it needs from remote servers when it needs it and can choose how to handle media (serve the original or cache and proxy). It already is similar-ish to a CDN.
AT-Proto is super complex, but my understanding is that a new server (app in AT-Proto parlance) needs to copy everything beforehand from all others, and needs to constantly replicate everything, wether it will be served or not, making the data transfers intractably massive.
Because one costs money and another doesn't. Simple fact.
CDN distribution of content is 2X the cost of static hosted files. This isn't a pendant saying "I CAN DO THIS" scenario, it's "can it be monetized", and in the case of of a video service on AT, absolutely not. Who do you think is paying for the hosting costs of a popular video in this scenario?
Cuban doesn't know WTF he's talking about about at all, but if he wants to launch competition and pay for that, there is certainly an expectation that a return will be built. Ads all over the place.