There are several issues here: finding, identity, storage, and delivery. They're somewhat separable.
Finding is "where do I go to get X". We have URLs, which assume a specific server. There are content-based keys, such as DOIs and URIs and hashes. But how do you find where the info is stored? Google? Something like DNS? Something else? This is the hardest problem. What Youtube really sells is "discovery" not streaming hosting. There are lots of streaming services, but you won't get the views.
Identity has all the usual problems. If people can create lots of identities at low cost, there will be spam and worse. No good answers there. China has this fixed, but they don't do anonymity. You need a government ID to connect to anything.
Storage is the big cost problem. Where does all this stuff go, and who pays for it? IPFS was supposed to be the distributed answer to this, and Filecoin was supposed to be the way to pay for it. That didn't work out. On the other hand, if someone wants it out there, then maybe they have to pay to store one copy.
Delivery can be distributed, but do you want to? Bittorrent was the prototype. Peertube is another peer to peer way to do it. Each video has a home server, and large numbers of people watching the same thing won't overload it because anyone watching the video also serves it. It works OK but is not as smooth an experience as YouTube. Plus it runs down your battery and runs up your bandwidth usage. Bandwidth is much cheaper in data center bulk than out at the end of a cell connection. Maybe do something like that but with ISP level caching servers, all serving each other. Sort of like Cloudflare / Akamai.
> Storage is the big cost problem. Where does all this stuff go, and who pays for it? IPFS was supposed to be the distributed answer to this, and Filecoin was supposed to be the way to pay for it. That didn't work out. On the other hand, if someone wants it out there, then maybe they have to pay to store one copy.
> That didn't work out.
Can you expand on this point? At my end, IPFS runs along just fine: My IPFS instance in particular continues to serve its content without much fuss. For Filecoin, the content hosted there is still on the network, there's just much less speculative fervor surrounding it & other networks like it.
Thank you, this is constructive. Let me complement with some more details:
Privacy. Messages in a private chat or group should not be visible to the operator or others. This likely necessitates e2ee, but it's possible that some multi-party protocol could protect from malicious operators without going full on e2ee (difficult on web).
Integrity: you don't want rogue operators to be able to falsify content. While you can't stop any website from faking content, you can stop it from propagating by having signatures and verifying them.
Availability: you should be able to post and fetch at all times, which necessitates aggressive caching. Signatures (from above) are useful here, because you can use untrusted storage layers.
Security: how do you prevent an instance operator from stealing your account or posting on your behalf.
Abuse: no system is anywhere near perfect, but the ones that come close are extremely centralized and use fingerprinting and behavior analysis. We probably don't want that, so what else?
Note that all of these are deeply tied into the identity problem. We would really need a form of PKI for real people (but without exposing real world identity). If people actually control their private keys, all of the above is greatly simplified. But even in defi, where custody of your private keys have a HUGE direct monetary value, most people don't want/know/care to set it up. So we have huge challenges.
Finding is "where do I go to get X". We have URLs, which assume a specific server. There are content-based keys, such as DOIs and URIs and hashes. But how do you find where the info is stored? Google? Something like DNS? Something else? This is the hardest problem. What Youtube really sells is "discovery" not streaming hosting. There are lots of streaming services, but you won't get the views.
Identity has all the usual problems. If people can create lots of identities at low cost, there will be spam and worse. No good answers there. China has this fixed, but they don't do anonymity. You need a government ID to connect to anything.
Storage is the big cost problem. Where does all this stuff go, and who pays for it? IPFS was supposed to be the distributed answer to this, and Filecoin was supposed to be the way to pay for it. That didn't work out. On the other hand, if someone wants it out there, then maybe they have to pay to store one copy.
Delivery can be distributed, but do you want to? Bittorrent was the prototype. Peertube is another peer to peer way to do it. Each video has a home server, and large numbers of people watching the same thing won't overload it because anyone watching the video also serves it. It works OK but is not as smooth an experience as YouTube. Plus it runs down your battery and runs up your bandwidth usage. Bandwidth is much cheaper in data center bulk than out at the end of a cell connection. Maybe do something like that but with ISP level caching servers, all serving each other. Sort of like Cloudflare / Akamai.