Actually it is a reasonable description of the problem. Domain names exist in 3 basic forms:

domain.TLD       where TLD (Top Level Domain) is com, net, org etc
domain.ccTLD     where ccTLD (Country Code TLD) is say .de
domain.SLD.ccTLD where SLD.ccTLD (Second Level Domain ccTLD) is com.au
+, co.uk, etc

# additionally we may have as a prefix
www.domain....
subdomain.domain.....
subsubdomain.subdomain.domain.....
[download]

So the task is how to extract the DOMAIN component +/- subdomains. The issue is that if you split on the . then the DOMAIN component has an index of either 0,1,2...N if you look from the left or -3 or -2 if you look from the right. See my post below for a heuristic that deals with most of the weirdness. The most annoying freature is that domian.ccTLD exists. If not all you would need to do is pop 1 element off if it was a TLD and 2 if it was a ccTLD. Alas that would be too easy!