On Sat, Mar 7, 2009 at 6:29 PM, Matt Warden <mwarden at gmail.com> wrote: > original table. *IF* your search words are always English words, then > you could significantly reduce the count of your substrings by > removing any that are not found in a set of English dictionary words. > This would probably knock out 75% of your rows and leave you with > essentially an index of all English words mapped to your domain names, > which is your ideal situation. > > Quite a bit of work, and depending on your requirements, may not even > be adequate. But just some ideas to mull over... If you go with this last option, I should mention you first have to generate all permutations, not simply each successive substring. So rather than simply: domain, omain, main, in; you would need to generate: domain, omain, main, in, domai, doma, dom, do, omai, oma, om, mai, ma. Then only could you cross reference with an English dictionary and eliminate rows not found. -- Matt Warden Cincinnati, OH, USA http://mattwarden.com This email proudly and graciously contributes to entropy.