Make the law clear and some defendant will get off on a technicality, but only once, then you fix the loophole. Over time the law evolves to more accurately reflect congressional intent as they pass patches to fix bugs, instead of leaving it to the courts and the biases (like prosecutors choosing who to prosecute when they want a precedent) that come along with that.
It's for this reason that, after significant wavering, I've supported the petition for the removal of Carmen Ortiz, despite strongly supporting her prosecution of other white collar crimes. She simply lacks appropriate discretion for the position, especially in light of JSTOR and MIT's not pursuing independent legal action.
At the very least it shouldn't be so easy to, for example, drum up 35 years worth of charges on excessive downloading.
"Fewer than 10 percent of the criminal cases brought by the
federal government each year are actually tried before
juries with all of the accompanying procedural safeguards
noted above. More than 90% of the criminal cases in America
are never tried, much less proven, to juries. The
overwhelming majority of individuals who are accused of
crime forgo their constitutional rights and plead guilty."[1]
From what I've read elsewhere, of the 10% or so that do go to trial, something like 9 times out of 10 the verdict is guilty. Given that a federal criminal trial takes $1.5 million to defend and that the prosecutor has all the weight and resources of the federal government behind then and the citizen has to rely on his own financial resources (which are likely to be frozen if they are well-heeled defendants), then that rate of conviction isn't surprising in the least. Federal criminal cases are totally asymmetric in favor of the federal government. As citizens, we always complain of well funded people prevailing in civil and criminal cases because of the money they have to defend themselves, but we ignore the well funded prosecution that US district attorneys have that allow them to consistently find people guilty who may not be.[1] http://www.cato.org/sites/cato.org/files/serials/files/regul...