From @mattseh on Sun Aug 14 2016 10:56:27 GMT+0000 (UTC)
If a single block in a 1GB file is changed, to update, a node only needs to re-download one small block, which is fine.
If a few bytes are added to the start of the file, or if a bunch of small changes are made, is this currently inefficient?
If so, is the support of diff objects planned? For example: ipfs_object_hash_1 + ipfs_diff_object_hash -> ipfs_object_hash_2
Copied from original issue: https://github.com/ipfs/faq/issues/159