Tilting at Windmills: Youtube case

So I have this insane project in my head and I need people to challenge it. As a lot of people in geminispace have technical background or ability this seems to be an appropriate place.

Problem Definition

I like watching videos on Yutube as it gives creatives a convinient video hosting platform, while users have search, subscriptions and discussion platform. It has incredible educational value, artisting value and part of human wealth of knowledge.

I hate Yutube/Google for having control over largest repository of videos. They gatekeep it with ads, regional locks and at freedom to remove them at any point to save on storage fees. There are smaller alternitives, but they are just clones of the same big ugly thing.

What I want is a free (no platform ads) publicly accessible video hosting platform alike wikipedia. But there is nothing like this, aparat from some federated fragmented services that barely have any videos and don't have great public uploading UX.

Solution Design

So. How could it be possible to take on Youtube and have any chance to win? This is a battle plan that starts forming in my head and I can't stop it!

1. Problem: Youtube network effect: huge video collection.

Solution: While we can't legally take all uploaded videos, we can take all CC videos as long as we credit all creators which we want to do anyway for ethical reasons. In 2012 there were 4 Million CC videos, in 2017 was 49 million. We can assume that there are 200-300 million CC videos right now. Even a part of those would be a great starting point.

2. Problem: Youtube is familiar user experience, people like what they already know.

Solution: Focus on providing better user experience - no ads, useful search, ability to download videos.

3. Problem: Storage costs.

Solution: Limit resolution to 720p max, start with videos over 100k views and less that 1h long. Average YT videos are 7-15 minutes long. With proper compression we can allocate ~20mb per video. 1 million videos would roughly be 20 Tb. Then AWS S3 storage per 1 Tb is $23-$1/month, depends on frequency of access. So tracking usage, starting with cheapest ($1) storage and moving videos into faster access storage as demand increase could allow to keep storage costs under $1,000 per month (in theory)

4. Problem: Traffic costs.

Solution: No users - no traffic. Jokes aside though, moving in demand videos onto AWS R2 with free outgoing traffic might be most cost effective solution. For low demand video CloudFront CDN directly from S3 might suffice. This is the fuzziest part of the plan.

5. Problem: Funding.

Solution: CPU, Storage and Traffic aint free. Without ads donations and paid b2b features look like only options available.

Please poke holes in the above plan! I'm already running one 1 million views per month website for $0/month, but the video component makes this one A LOT more complex to implement and balance costs. Could it be that video hosting of such scale can only be commercial? I'd love to believe that's not true.

Nov 28 · 2 weeks ago

5 Comments ↓

It is a tough nut to crack. I don't think it's possible without bullshit financial crap.

I think it´s important to recognize that multinationals like Google do not play by the same rules of business as we do. That is, they do not have to balance their books for individual products, or have their products actually generate cash. People are not their customers -- money often comes from complicated exchanges with other businesses or leverage of customer base in other products. The bulk of their operating funds comes from their unlimited ability to borrow and often straight from the government. And given their wingspan they have an unlimited ability to create structured financial arrangements. They can run a losing business for decades and undercut and sue you until you go bankrupt or die.

In the meantime you will be faced with very real issues of finding capital for the infrastructure, or having to rent it from Google or Amazon!

You could try some distributed network where storage is shared, but getting reasonable streaming is likely difficult that way due to unpredictability of such infrastructures, and managing millions of videos is expensive enough without some of them disappearing and reappearing.

I have to admit that I admire the Youtube algorithms. They are incredibly fast at identifying my fears and desires and engaging me almost instantly. They also very quickly figure out who I am on different machines and geographic locations. I think it takes maybe 3 searches and I have my favorite topic popping up as if I am home.

None of the alternative viewers like Grayjay are anything like that -- what they feed me is annoying and I get tired of looking at the handful of things I explicitly subscribe to or search. Youtube is just so much better at providing me with new 'discoveries'...

There is a people component you are not considering. In a free service where users may post, you invite the very worst of humanity to see what they can get away with. Who will moderate that and how? Without moderation, the best you can hope for is tons of pornography and copyrighted material being uploaded en masse. This presents a liability problem to the platform as you must play whack a mole with every new copyright violation and challenge the age of the performers. Wikipedia has a dedicated lawyer they pay to handle this and the volunteer editors are the first line of defense. I applaud the Athenian ideals but the target species is driven mostly by id when given anonymity.

That is indeed and obvious aspect that I haven't considered at all, focusing mostly on technical aspects. My first thought is that volunteer moderators might be a possible approach. It works for wiki, and for some community sites like reddit. Something to think about for sure.

Something you're missing is that people post on YouTube because YouTube pays them. If you want to snatch creators from YouTube, you'll have to pay them competitively.

There have been attempts, though the project names escape my memory, to try to use torrenting or similar p2p file sharing to reduce storage and traffic cost. The idea is usually some play on users seeding videos after watching them. But that has privacy implications (3rd parties can easily monitor whose watching what), among other issues.

ISPs can get annoyed with you if you're seeding torrents under the assumption it is piracy, ignoring the perfectly legitimate uses such as helping to distribute Linux ISOs.