> Why do you mention "custom collations"? Because any check for equivalence, which would be required for the hypothetical "was it *really* modified?" check, must (it seems to me) honor column-level collation (if any). Without that, "a" and "A" would be different values even if the collation is "nocase". If that collation is not honored, some percentage of users would rightfully point out that "a" and "A" are, for the "nocase" collation, equivalent, and therefore (under this hypothetical feature) should not qualify as a change. Others would, also arguably rightfully, beg to differ because the binary values differ. That said, that's all based on deduction, not any familiarity at all with the code, and therefore may be completely wrong.