Like others have said, it’s way more resource intensive than text based systems. Even discounting higher res vids, if you go to any random larger YT channel and download all their videos in 144p 480p and 720p it’ll be quite a lot larger than you might expect. Sure, if you’re serious about it you could get an array of hard drives and a small server, but you’re talking hundreds of bucks and lots of upkeep. Outsource it to a VPS and AWS buckets and you’ve still got upkeep but now you’ve added an extra 0 to your bill.
There’s not enough charitable nerds on the internet to host even a fraction of 1% of Youtube. It’s even worse if self hosting instances is pushed. Even as a fellow tech nerd, no way I’m hosting my own instance just so I can share a video once in a blue moon. Something that always gets my goat in fediverse discourse is when people always jump to saying something along the lines of “just host your own” then wonder why AP went from ~2.5M users to 0.8M users.
There’s also some Fediverse specific issues that hold back a more mainstream audience. There’s some fringe political stuff on both sides of the isle which can pretty easily scare off people, and defederation combined with peertube’s more siloed approach makes discovery near nil. (can’t see content from remote peertube instances unless somebody has already subscribed to that channel on the remote instance from your local instance AFAIK).
Then there’s the new platform (or in this case many platforms connected via one protocol) issues. Lack of users, limited/no monetization, limited development/support, and very few pros + a lot of cons at first glance from somebody who doesn’t consider tech a hobby and is comparing it to established platforms.
Edit: Can’t remember who, but iirc a peertube user I follow who regularly deletes their videos because their host doesn’t give them too much space. It’s great for a less big tech way to see their latest videos, but not acceptable if anyone’s gonna bill something like that as the next big video platform.
Looks like a cross between Zot and Nostr. It looks like it’s got some interesting features, namely the key based identities, easy migration, and it being designed around implementing custom features into the protocol while hopefully remaining compatible with other poly servers.
But I’m not really sure there’d be too much of a point in using it. Zot, Nostr, and AT (Bluesky) both already have the former two - with Nostr having the latter as well, and AT having a limited amount of the latter (Lexicons add flexibility but break compatibility). Although I only skimmed through the docs I can’t see features big enough that would make someone implement polyproto over using an existing protocol that already has a userbase.
It might be handy as a server that is centered around polyproto, but compatible with other software (kind of like how Friendica isn’t actually activity pub, but bridges things internally). That way you’d get the benefits of polyproto when talking with others on poly, while still being able to communicate with others on other protocols. But it doesn’t look like they’re aiming for that - I’m not a cryptography expert, but they’re using X.509-based keys. Both Nostr and AT use secp256k1, so had they used those you’d be able to use the same identity across multiple platforms (even if the posts had to be bridged/translated since the json formatting would be different).
If you’re looking for a central home server, private key powered accounts hidden behind a domain based username, the ability to instantly migrate servers, and a protocol that supported heavy customization, I’d recommend looking into ditto. It’s built on Nostr, but it’s built around being a more Activity Pub style home server (and it even works with Mastodon API capable clients). Ditto is nearly 1:1 with what it looks like polyproto wants to build, and it already exists right now on a protocol with an active userbase, so I’m not quite sure what the goal would be in making another protocol.
Or maybe I’m just being dumb and there’s more to polyproto than I realize.