Matching text nodes with only whitespaces

Hi,

I need to write a template matching text nodes, which just consit of
multiple whitespaces (tabs in most cases).
I cant't use normalize-space since I need leading and tailing
whitespaces in some cases.
I suppose matches() would help me here, but I actually don't know how to
formulate the regular expression for that.
Can you help me here?
Thanks!

wbr,
Roman