How the comparison works
The tool computes a longest common subsequence between the two texts — the same approach git diff uses. That matters: comparing the two line by line in parallel would report every line after a single inserted line as changed, which for a diff is not an approximation but the wrong answer.
Finding the longest common subsequence means the tool identifies which lines genuinely correspond between the two versions, so an insertion shows as one added line rather than shifting everything below it.
Ignore options
Ignore whitespace collapses runs of spaces and trims the ends of lines before comparing. Use it when re-indentation or a trailing space is hiding the real change.
Ignore case compares case-insensitively. Useful for prose where a capitalisation change is not the difference you are looking for.
Both change what counts as "the same line", so the diff will look different with them on. The displayed text is always the original, with its original casing and spacing.
Size limit
The comparison is capped at 2,000 lines per side. The algorithm builds a table proportional to the two lengths multiplied together, and past that point the browser tab would lock up rather than finish. If you are comparing something larger, a command-line diff is the right tool.