Suggestions for a slight revision to the substring proposal to reduce
the number of special cases:

* The positions should be 1 based (as much as I'm a zero based person)
  - It's more consistent with result sets being 1 based.
  - It allows for:

* Both start and end position should allow for negative, 0, or positive
integers. Negative or 0 means count back from the end that many
characters.

eg:

substring="1:3" -- the first two characters
substring="1:0" -- the entire string
substring="1:-1" -- entire string minus last character
substring="-2:0" -- the last two characters // not possible currently
substring="2:-1" -- the string minus first and last characters
substring="1:1" -- the first character
substring="0:0" -- the last character

Rob