Knowing without evidence

Ruth Marcus' column yesterday brings up a number of good points and for the most part, she's on the money. But there's a section that makes no sense at all:

The difficulty lies in how to deal with suspected terrorists at the outset. The Obama administration no doubt would recoil at the back-to-the-Bush-administration atmospherics of indefinitely detaining an enemy combatant on American soil. But what to do in cases like that of Padilla, where the evidence is still sketchy but the suspect too dangerous to remain loose?

Work with me here, Ruth. If the evidence is all that "sketchy," if it's just not possible to point to anything specific to determine that someone should be detained, then how do you know they're too dangerous to let loose? Or are you in favor of allowing detention on the basis of hunches? Well, dude looks like he might do something, so we're going to hold him just in case.

She goes on to discuss one option--allowing detention up to 14 days without judicial approval--but that doesn't address the problem. At the end of 14 days, you've either got evidence or it's still 'sketchy.' Would she still somehow magically just know which people were too dangerous to be let loose? How? It's not like prosecutors have a hard time finding sympathetic judges willing to authorize detention even in marginal cases on national-security issues.

Yes, there are gray areas. But the entire assumption that we should be holding people "just in case" is a threat to the rule of law, which has taken (and is still taking) a beating.