Right now the cooked index does case-insensitive comparisons. But is this really needed? I think there are a few case-insensitive languages in gdb. GNAT already case-folds all symbol names. In fact, due to the way GNAT encodes non-ASCII characters, case-folding in the index might mean that you could construct a situation where symbol lookup finds the wrong thing. I haven't looked to see what Pascal, Modula-2, or Fortran do. However, for those, one option might be to do the folding as a canonicalization step.
Actually, this is needed to handle "set case-sensitive off". It might also be needed so that Ada users can print C globals without trying to get the case correct (though in a brief test I couldn't get that working).