Rethink replay protection

[ethanfrey]

Replay protection is super important in a blockchain.

Right now, Basecoin uses sequence to work as a nonce. However this requires all TX to have a Basecoin input.

Can we think of a form of Replay protection that works when not all TX have a Basecoin input.

[ethanfrey]

One idea is to make it a generic layer after the sig checking, and use the tx hash of the signed data as a nonce (so you can't just add an extra sig to change it). If it is successfully DeliverTxd, then store that hash in a lookup table, and reject any tx that is in that table.

This would work without each app having to maintain it's own counts (like coin sequence) and the client to remember them all. Also if you have a few outstanding multisig tx, then the sequence number can be rather limiting (okay, i can't send the multisig with you until the other multisig is completed). However, it would also require ~20 bytes per tx, which may grow quite large if we have 10tx/s sustained (-> 860K/day, 300M/year = 6-7GB/year)

Not sure where to find the balance between usability and performance.

[ethanfrey]

I think those numbers 6-7GB/year in a very active chain are not actually that bad - we will probably have much more state than that. Let's go for it. We can optimize later by adding a bloom filter.

The changes would be:

1. Add Nonce int32 to ChainTx, which is filled with a random int on NewChain. This allows the same tx to be intentionally sent twice (Ethan sends Rigel 5 axi every week).

2. Add the Replay middleware that checks on the hash of the tx bytes, and test it. Also add some benchmarks. (Add some helper function for tx.Bytes() and tx.Hash() that we can also later add to the codegen).

3. Add the replay middleware in the default stack, right after the Signature middleware

4. Remove the use of sequence from coins and all the cli as well, as they are no longer needed.

Anything else? Looking forward to a PR, rige :)

[ethanfrey]

Okay, this is the same as #160, which is a more clear explanation. Keeping this for history... maybe there were some good ideas here...