This is looking pretty good. I’ll raise one issue here that may be worth thinking about(although since the items are all vecs it may not be an issue.
Do we need to review if any of the Value types or Error types need to be opt Value or opt XXX_Error for reasons expressed in this thread: ICDevs.org Voting Incident Report and Stable API Discussion
Are there any places where this would make pseudo-compatibility with an evolution of this standard easier? The result would be that nulls would be returned in place of future Value(ICRC16) or XXX_Error(If we want to add errors in ICRC8 or some other standard) that don’t match the exact variants offered without breaking compatibility. Canisters can choose how to handle unknown nulls.
I may be overthinking this. Perhaps the NNS team can weigh in here with an opinion? cc @sasuki