str:tokenize() and consecutive delimiters in the source string

Hi list,

what would be the expected return value of str:tokenize('a, b',' ,') ?
Some processors return
  <node>a</node><node>b</node>
while others, for example libxslt, return
  <node>a</node><node></node><node>b</node>
i.e., there is an empty string seen between the comma and the blank. From the 
spec it is not clear to me, whether consecutive delimiters should behave like 
one single delimiter, or not.

An even simpler example would be str:tokenize('a  b',' '), with two blanks 
between a and b. The function name suggests that tokens are returned, so 
personally, I would expect to get two tokens (i.e. whitespace collapsed) and 
not three. But to sue libxslt, the spec should imho be more precise, or there 
should be an example given, similar to one of the two above.

Thomas