Require lexical order for Context Inputs

tips/TIP-0045/tip-0045.md

@@ -1185,9 +1185,12 @@ The following criteria defines whether a payload passes the syntactical validati
   - Context Inputs:
     - `Context Inputs Count` must be `0 ≤ x ≤ Max Inputs Count`.
     - `Context Input Type` must denote an allowed input type, listed in section [Context Inputs](#context-inputs).
+    - The Context Inputs must be lexically ordered and unique, both based on the following comparison criteria:
+      - `Context Input Type` as the first criteria for all Context Inputs.
+      - `Index` as the second criteria for _Reward Inputs_.
+      - `Account ID` as the second criteria for _Block Issuance Credit Inputs_.
     - There must be zero or one _Commitment Input_.
-    - All _Reward Inputs_ must reference a different `Index` and it must hold that: `Index < Inputs Count`.
-    - All _Block Issuance Credit Inputs_ must reference a different `Account ID`.
+    - For each `Index` in the _Reward Inputs_ it must hold that: `Index < Inputs Count`.
   - Outputs:
     - `Outputs Count` must be `0 < x ≤ Max Outputs Count`.
     - For each output the following must be true:
@@ -1268,6 +1271,12 @@ used deposit address. Ed25519 addresses are not meant to be used like email addr
   side mismatch. While this would reduce the size of a transaction, it would make it impossible to debug the reason for
   having an invalid signature and transaction. With the current solution we intend to optimize for ease of development.
 - Uniqueness of all inputs is kept as it prevents introducing double spends in the same transaction.
+- The sorting criteria for _Commitment Inputs_ is only based on their _Context Input Type_ and does not take into
+  account the _Commitment ID_, which allows for an optimization when determining whether there are zero or one
+  _Commitment Inputs_. An algorithm checking lexical order and uniqueness will detect the violation of the desired "at
+  most one" semantic as soon as two or more _Commitment Inputs_ are present, since those would be considered duplicates.
+- Requiring lexical order in various places (such as for _Context Inputs_) is mostly an optimization in order to allow
+  for efficient duplicate detection.
 
 # Test Vectors