Good Idea from Mr. Nakamoto:

Whenever a new block is accepted, the best block pointer is updated in blkindex.dat.  If you also record it in wallet.dat (as a CBlockLocator), then on startup, if wallet's last seen best block is further back than blkindex.dat's, it should rescan starting from there.

Good idea. What happens when wallet's block is from an orphaned chain (chainsplit while wallet was closed)?

CBlockLocator Does The Right Thing. Good idea to test that, though:

• Test wallet.dat points to orphan block that IS in blkindex (should rescan starting at block split)
• Test wallet.dat points to orphan block that IS NOT in blkindex (should do full rescan)

Now we just need a good way of forcing chain splits / orphan blocks on the testnet...

Forcing chain splits with two GPUs and -addnode/-connect/-nolisten magic should be easy. I can do that once we have a patch.

See #152

Is -rescan needed because of Merkle tree pruning? Where I live disks are cheap but the network is prohibitively expense. Is it practical to optionally save the ENTIRE block chain and just look up my addresses in the full local table.

Or similarly, is it possible to keep track of all addresses ever seen on a local machine and never prune those addresses, regardless of the 'current' wallet?

-rescan is to find transactions that are in your local copy of the block chain, but not in your wallet, yet should be.

Also, merged since 0.3.21.

Tested on seven wallet.dat's that have not been opened in a week. Confirmed:

1 That 0.3.21 does indeed rescan (without the -rescan flag)

2 That each subsequently opened wallet requires little or no bandwidth

3 Possible to open wallet and view balance without network connection

As for # 3, I wonder if there should not be a message in the GUI indicating that the balance is accurate only as of some date and hint that an internet connection is required to get an up-to-date balance.

• Note one week (May 2011) required almost 20MB of downstream bandwidth.