Comment

Comment on Cloudflare is bad. Youre right.

ramble81@lemm.ee ⁨6⁩ ⁨months⁩ ago

The bigger trouble is creating a CDN has a stupidly high barrier to entry. You literally need your own data centers across the world, your own server infrastructure, the man power to manage it, etc.

You could try to host it on a cloud provider but you’d go bankrupt even quicker. Unless someone were to try to build a co-op run CDN, it’s just not gonna happen without a profit motive and a large amount of capital.

source

Sort:hotnew top

vzq@lemmy.blahaj.zone ⁨6⁩ ⁨months⁩ ago
That’s true. The bizarre paradox of the centralization of edge infrastructure is real.

source
yannic@lemmy.ca ⁨6⁩ ⁨months⁩ ago
I once realized so many of my favourite businesses were cooperatives. I started thinking of what other co-ops I could start and grow. The excitement faded once I realized it would have to not be about the money.

source
- laughterlaughter@lemmy.world ⁨6⁩ ⁨months⁩ ago
  Car making without the tracking bullshit!
  
  source
NotMyOldRedditName@lemmy.world ⁨6⁩ ⁨months⁩ ago
I feel like something like www.storj.io is on the path to what we would want/need?

There might be some additional requirements for a true CDN to ensure data is closer to where it’s needed and in as many regions as needed though.

source
muntedcrocodile@lemm.ee ⁨6⁩ ⁨months⁩ ago
I mean the optimal cdn is maximally distributed to reduce load and latency right. Unfortunatly the web was not built in a manner that supports this.

Eg if we could have a single url for the same object that could be served by any server that is part of the fediverse then the fediverse itself would be an optimal cdn.

Perhaps we should take some notes from peertube.

source
- ramble81@lemm.ee ⁨6⁩ ⁨months⁩ ago
  Look up Anycast when you get a chance.
  
  source
  - muntedcrocodile@lemm.ee ⁨6⁩ ⁨months⁩ ago
    I consulted with professor gpt and it seams that it’s basicly just giving the same ip address to multiple servers meaning that any of said servers can serve as that ip.
    
    Also it seems said ips require paying large sums of money to isps. My poiny was more that with the current mainstream internet (http websockets etc) it would require you to run a local service/proxy that can interpret a global id and route to basicly any small server with said resource. Unfortunatly i dont think its possible to build such a thing that would just work across browsers if embedded into a standard webpage.
    
    source