It reads "Splits str into a list of strings separated by matches of re" but the current reference implementation seems to "eat up" any empty strings between matches.  Suppose I want to parse text fields separated by a comma or a semicolon:

  (regexp-split '((",;")) "a,,b,")

I'd expect to have ("a" "" "" "b" "") rather than ("a" "b").

Note: If we allow empty strings between matches, we'd better to refer explicitly to the case when regexp matches an empty strings, since one can argue that such regexp splits an empty string to infinite number of empty strings.  

P.S. The link to the reference implementation seems obsolete (it points to code.google.com, and redirected to github project toppage but not the actual file.)