Re: [bitcoin-dev] [patch] Switching Bitcoin Core to sqlite db

[Peter R <peter_r@gmx.com>]

> On Sunday, November 15, 2015 1:02:33 AM Peter R via bitcoin-dev wrote:
>> A group of us have been exploring this “meta-cognition” idea with Bitcoin
>> Unlimited.  For example, Bitcoin Unlimited can be (optionally) made to
>> automatically fork to the longest chain if it “gets stuck” and can neither
>> prove that a block is valid nor that the block is invalid.
> 
> This situation isn't something that can be ignored and simply moved past. If 
> you can't determine the validity of a block, you also cannot process its 
> results correctly. Taking for example the BDB/LevelDB issue, the result was 
> that BDB failed to accept further changes to the UTXO set. Unless the UTXO set 
> could be updated correctly, there is no way to even attempt to validate the 
> next block or any new transactions.

Great point, Luke! 

Indeed, whether the program can or cannot continue after a Type 1 consensus mismatch depends on the specifics of the situation and exactly how the code was written.  But I agree: there are cases where the program *can’t* continue.  In those cases it would halt.  This would require manual intervention to fix but avoids the problem of potential double-spends during the fork event.  This would be preferable to knowingly causing a fork.  

Peter