Realization I just had: Of course, a distributed hashtable used with Gzz
should not only be able to store *peers* that host a given block (as in,
the peer hosting the block tells the DHT so), but also *regular HTTP
servers* (as in, a peer tells the DHT, "I know that block X/the body of
block X can be downloaded at URL Y).
AAAAAAAAAAAAGH! YES!
Absolutely. This puts things really nicely together.
For files that cannot be deeply linked, we can still have a reference
saying, 'you can download that block if you go to this web page,' though
this would need to be coupled with some way to assert credibility,
otherwise you could confuse people by directing them to arbitrary pages.
That's true for the above as well; you could otherwise DDOS someone
by putting their server down for all blocks in the known universe.