Give proof on a generalized watcher only needs to check basic things

[boolafish]

Note

The concept is that watcher only needs to check two things:

1. check sum(input) >= sum(output): no money out of the air.
2. output predicate: can the tx output be considered "used".

Successful criteria

• Provide some logic reasoning, and proof it should be secure enough.
• Document the reasoning down.

[boolafish]

To be more specified, this means we would have a "core watcher" that does the basic verification (predicate + input output sum) and can plugging extension logic for their own need (eg. DEX settlement logic verification).

Let's say we are plugging a new predicate new_P: Case 1: now you are creating a new tx that is totally un-related with the new_P. The history of the new tx you are creating have never touched new_P.

• apparently safe as it is totally unrelated. And even though they can exit, as long as they cannot exit multiple time, their exit would not impact your money.

Case 2: now you are creating a new tx that new_P was used in one or some of the inputs' history, but the new tx is not using new_P directly.

• This means the tx that uses new_P is already in history. That tx cannot exit anymore as it would be challenged by the predicate new_P. You are safe to create the new tx. Even such tx was invalid it become "valid" as a fact now. (In RC tx, it means if exchange + operator fraud, it is possible to steal fund from custody and no-one else would recognize it)

Case 3: now you are creating a new tx that actually needs to fulfill new_P.

• Then you would actually plugging the extension to verify fully as you cares. Then you are not running the the "core watcher" mode.

[boolafish]

@paulperegud @pik694 @pgebal Do you think the above statement sounds solid enough?