Merklizing the Key/Value Store for Fun and Profit
Metadata
Highlights
- Since a directory is identified by the hash of the serialized list of its entries, and each serialized entry has its own hash, we get a system where arbitrarily large directory trees can be uniquely identified with a single constant-size hash. These hashes allow IPFS to be “self-authenticating”: users request content by hash, and can simply hash the returned bytes themselves to ensure they’ve received what they expect. (View Highlight)
- Your friends use their merkle paths to prove inclusion in a set uniquely identified by a single constant-size hash, and it only takes O(log(n)) hashes instead of the O(n) table scan. (View Highlight)
- Why wouldn’t you just give security the whole list and ask them to build an index over it? Why not use a database? One reason might be because security is running entirely on a blockchain like Ethereum (don’t ask me what this means in the analogy) and storing data with them is super expensive. The merkle tree lets you compress the data you need to commit into a constant-size hash. The tradeoff is that your friends have to remember their merkle paths.
# (View Highlight)
New highlights added August 21, 2023 at 6:12 AM
- Overall it seems like there’s an emerging vision of a lower-level type of decentralization, where the data structures themselves sync directly with each other. The modern computer has hundreds of databases for various kinds of apps at all times - what if the application layer didn’t have to mediate sync and replication logic at all? What if you could make a social app without writing a single line of networking code? (View Highlight)
Merklizing the Key/Value Store for Fun and Profit
Metadata
Highlights
- Since a directory is identified by the hash of the serialized list of its entries, and each serialized entry has its own hash, we get a system where arbitrarily large directory trees can be uniquely identified with a single constant-size hash. These hashes allow IPFS to be “self-authenticating”: users request content by hash, and can simply hash the returned bytes themselves to ensure they’ve received what they expect. (View Highlight)
- Your friends use their merkle paths to prove inclusion in a set uniquely identified by a single constant-size hash, and it only takes O(log(n)) hashes instead of the O(n) table scan. (View Highlight)
- Why wouldn’t you just give security the whole list and ask them to build an index over it? Why not use a database? One reason might be because security is running entirely on a blockchain like Ethereum (don’t ask me what this means in the analogy) and storing data with them is super expensive. The merkle tree lets you compress the data you need to commit into a constant-size hash. The tradeoff is that your friends have to remember their merkle paths.
# (View Highlight)
- Overall it seems like there’s an emerging vision of a lower-level type of decentralization, where the data structures themselves sync directly with each other. The modern computer has hundreds of databases for various kinds of apps at all times - what if the application layer didn’t have to mediate sync and replication logic at all? What if you could make a social app without writing a single line of networking code? (View Highlight)