diff --git a/ansi-tests/TODO b/ansi-tests/TODO
index a19d9b29d3681d142916da0a6ff9d5e740a58271..7860fe010aff9ccfffd3f44ae0fc206c586b3bb4 100644
--- a/ansi-tests/TODO
+++ b/ansi-tests/TODO
@@ -80,3 +80,10 @@ Got:      0
 23. Modify rt so that when failing tests are reported, they are grouped by
   :notes and the :notes comment is printed out.  This will help explain
   what the failures mean.
+
+24. Add float/rational comparison tests where the rational is (exact rational for the
+  float) +- a very small rational.  This will check if the rule in the second paragraph
+  of 12.1.4.1 is being followed, or if the implementation is incorrectly floating
+  the rational then comparing.  Also add tests of very large (non-floatable) rationals
+  with floats, if they are not already present.
+