With most online services, some entity X (Google in this case) says 'I will host this data.' Eventually it stops being worth it for X to host that data, and the data is lost forever (unless someone archives it like this case, which just changes the X to some other entity).
With distributed content sharing networks, the data is retained as long as anyone is willing to bear the cost of storing and distributing it. After I upload a file I may eventually decide that it is no longer worth sharing, but as long as someone thinks it is a useful file to share they can have it seeding from their computer. The file is only permanently gone once nobody thinks it's worth sharing. And with a large population on these networks, that is equivalent to the file having no actual value - except for historical research perhaps.
This is actually better than what the parent (mavhc) suggested, because with the donation method there is still one entity in charge of sharing the data. Others can archive and share it (if allowed by the primary entity), but there is still a discoverability problem that is solved by IPFS/BT.
If ipfs suffers the same problem - noone knows which hash to look up - the files will be irrevocably lost.
-- Quinn Norton
https://medium.com/message/hello-future-pastebin-readers-39d...