> If someone is a known murderer, what would you do? You have to section that person off somehow.

That's a pretty simplistic way to look at it. Do you know how likely a known murderer is to commit any offense, much less another murder? Researchers spend a lot of effort into trying to figure that out, here's a quote[0] from one such paper:

> Our study indicates that it is a misconception that homicide offenders released after long-term incarceration are hardly ever involved in violent crimes upon release. This misconception stems from several studies that have found that homicide offenders rarely re-offend with a second homicide. After analyzing a large volume of data, our study found that none of the 336 released homicide offenders committed another homicide, but approximately one-third of both the felony homicide offenders and general altercation precipitated homicide offenders do re-offend with new violent or drug offenses.

> Additionally, homicide is not a homogeneous behavior. Homicide perpetrators are not the same in terms of motivation, environmental factors, demographics, and interpersonal dynamics. Different factors of complex combinations precipitate homicides that range from felony murders in the midst of an armed robbery to murders involving sex, love, and emotion, and murders for money and property to murders because of drug use and alcohol consumption

[0] - http://www.newjersey.gov/corrections/pdf/REU/Recidivism_Amon...

I don't profess to know the answer of the underlying question - how you should "deal" with criminals. But reading that study, none of the factors that lead to recidivism are treated in prison(alcohol, drug addiction), some are actually exacerbated(poor economical situation, low education).

You're conflating homicide and murder. Legally there are many types of homicides. The manslaughter types are already sentenced less because as you point out they are less likely to occur again. I agree with your points, but wanted to note that the law already has some framework to deal with the differences.

> re-offend with new violent or drug offenses.

Putting drug offenses in there kind of makes those numbers pointless.

"knowing" someone is a murderer is actually quite difficult with all the recent discoveries of "junk science" that have historically been used to prove guilt.

? What's your point?

His point is that a lot of people have been and keep being sent to prison thanks to DNA testing and other methods, recently proven to be unreliable.

Therefore falsifying the premise to the entire argument GP was replying to.