TODO

Very soon

• Check VM unit tests are good.
• VM unit tests should include calls.
• Full state unit tests.
• Tests for trie ref counting.

Fairly soon

• Include files in LLL.
• Compiler & code-gen library for LLL.

Soon

GUI

• Make address/block chain list model-based, JIT populated.
• Make everything else model-based.
• Qt/QML class.

Core

• Signalling for changes to State.

Thread-Safety

• BlockChain
• TransactionQueue
• State

Cleanups & caching

• All caches should flush unused data (I'm looking at you, BlockChain) to avoid memory overload.

Network:

• "Error sending: Broken pipe" should only display once before killing session.
• Genesis block to encode network/PoC version.

Soonish

Crypto stuff:

• kFromMessage
• Check all the tweak instructions.

Network:

• *** Exponential backoff on bad connection.
• *** Handle exception when no network.
• *** Only download blocks from one peer at once.
  • Parallelise block download.
• Better handling of bad blocks
  • Track which peers passed which blocks and punish them.
  • Don't pass on block until they're known good.
• NotInChain will be very bad for new peers - it'll run through until the genesis.
  • Check how many it has first.
• Crypto on network - use id as public key?
• Make work with IPv6
• Peers rated.
  • Useful/useless - new blocks/transactions or useful peers?
  • Solid communications?
• Strategy for peer suggestion?

Cleanups & caching

• State DB should keep only last few N blocks worth of nodes (except for restore points - configurable, defaults to every 30000th block - all blocks that are restore points should be stored so their stateRoots are known good).
• Logger: cleanup windows-specific code, determine impact of Release vs Debug, settings

General:

• Better logging.
  • Colours.
  • Move over to new system.

Robustness

• Remove aborts
• Recover from all exceptions.
  • Especially RLP & other I/O.
• RLP should never assert; only throw.
• Better handling of corrupt blocks.
  • Kill DB & restart.

GUI:

• Turn on/off debug channels.

Later

Trie on DB.

• Move the restore point stuff into block restore points
  • i.e. keep all nodes from last 127 blocks with counter, at 128, kill but keep every (60247)th or so i.e. one per week as a restore point.
  • maybe allow this to be configured.