Add prefix for block / transaction

[rabbitprincess]

Trie key is saved with 's' prefix in #200. it can prevent overwritten by contract values.
But, I think that is not sufficient, because block and tx also saved without prefix.
to prevent overwritten, block and transaction also need to add prefix when saved in db.

[kroggen]

As long as all content that is stored using hash as the key has this format below, there is no risk of conflict:

hash(linear content) -> content

It only happens with the merkle trie because the storage format is different:

hash(first 2 children) -> content

So if the hash of block and tx are computed using "linear content", there is no risk of a hash to map to a different content

[kroggen]

Indeed. I have not thought that the state data is stored on a separate database.
That is an important information. Good catch!

[kroggen]

The STATE_DB_KEY_PREFIX is added to the key, and then the key is hashed on the state trie. So that prefix is not stored on the database. It may be there to avoid collision with other internal keys that can be stored on the same state trie.

#200 may still be required, because the problem is not with the key but with the value. They are only converted to JSON, so strings are stored with quotes

[rabbitprincess]

i agree. #200 should be merged. but before that, we need to prepare for db migration.

[rabbitprincess]

and maybe, i can find any ideas to tidy database size. ethereum has x300 more transactions than aergo, but eth`s state size is not large. ( aergo node size is 1 TB, and ethereum full node is around 12TB )

[kroggen]

The focus of development of Aergo was speed over used disk space. It appears there is a trade off between the 2. The "trie node batches" makes it faster on read and write operations (less I/O) but uses more disk space. I don't now if there is any other thing that also affects the high disk usage.

But again, let's focus on the security issues, dependencies and the release of V4 now.