> But if the live mirrors go away and the hash method or chunking method change (and they do) then the hash is permanently useless
No, it might be permanently useless but it could also become useful in the future. Not only could you ask around archives if they have that hash, some random peer might also appear with the content in the future. Although unlikely. Although infinitely better than location-addressing that will for sure break at one point.
> so having the thing the hash was generated from isn't enough to make the hash live again
Of course, if it was that easy it wouldn't be a hash :)
> But that's just IPFS... we really deserve a better content address system
I was not considering IPFS, just thinking about content-addressing in general.
No, it might be permanently useless but it could also become useful in the future. Not only could you ask around archives if they have that hash, some random peer might also appear with the content in the future. Although unlikely. Although infinitely better than location-addressing that will for sure break at one point.
> so having the thing the hash was generated from isn't enough to make the hash live again
Of course, if it was that easy it wouldn't be a hash :)
> But that's just IPFS... we really deserve a better content address system
I was not considering IPFS, just thinking about content-addressing in general.