A short, excellent article; full title being 'U.S. Strategy for Countering Violent Extremism: An Assessment' by Clint Watts (aka CWOT) and Will McCants:http://www.fpri.org/articles/2012/12...ism-assessment

They end with:
we ask the U.S. government to take a hard look at the number of people who support terrorist groups. If their numbers are small and if their violent acts are few, a traditional law enforcement approach might be all that is required. We know that CVE is viewed by some as a more holistic approach to terrorism. But the downside of holistic approaches is that they can do a whole lot of harm and consume a disproportionate amount of resources. If CVE is required, identify sympathizers and supporters, select a limited set of actions, execute them through a few agencies and measure their effectiveness against defined objectives. Such measures might prevent a few people from pursuing a bad course in life, and if the measures fail, law enforcement knows what to do.
Interesting to note how the UK experience in this field has had so much effect, a classic case of "selling" a bad product.