How do dat:// sites interact with servers? (2018)
pfrazee.hashbase.io
pfrazee.hashbase.io
And here's a walkthrough of building a simple chat application: https://docs.dat.foundation/docs/kappa
It seems like peer-to-peer should imply that you can connect to mirrors without trusting them? Any data you download gets verified. But I guess that means you have to trust that there are no security bugs in your client, or eventually you'll hit a server that exploits it.
In practice, I think picking one caching server might be more reliable. (Much like setting GOCACHE when fetching Go modules, or configuring DNS with a list of servers.) We should be aiming for making caching servers interchangeable, but not necessarily connecting to random servers all the time.
This moves the problem one step back, since the caching servers will be connecting all over the place (much like recursive DNS), but at least it's running in the cloud with better connectivity.