As a CS professor, I can't tell you how many times we've lost students with great potential in CS because they had no prior experience but were comparing themselves to inferior students with a year or two of programming experience in high school. If you get the students who have prior experience into a "fast track" class (e.g. that compresses the first year into a single term) then both the "experienced" and "naive" students can actually learn at their own pace. Fortunately, I teach at a small college, and so most times we can identify those students and get them into a better class. And I'm actually in favor of having students with a lot of experience start by skipping a class or two. The sooner students are surrounded by their "peers" in ability/experience, the faster and more reliably they're going to engage.
But to be clear: the issue isn't that people should be actively sorting the students so that only female and non-white students are in the CS1 class. That's a horrible idea, racist, sexist, and all the other "ists" you can come up with. It is likely that the "normal" track will have more non-white and female students in it because that's what the high school demographics say: non-white/non-Asian/female students are less likely to have prior experience. But it's also true that there will be more students from rural schools in the "normal" track, because rural schools are less likely to have computer programming courses.
This is a larger problem than in Ohio. In Montana there are a small number of Amish and various other Anabaptists (all of which consider judicial action "taboo"), and they're also finding themselves square in the crosshairs. The fact is that Anabaptists tend to choose to live in isolated areas (so they will be left alone), yet those isolated areas are the ones that are increasingly being exploited for natural resources.
It's also important to understand some of the other restrictions that aren't obvious. If an "English" farmer has a railroad that is forced on him/her through his/her property, s/he can request a crossing be built so that the normal operations of the farm (like moving cattle) aren't impeded. But to do that requires the farmer carry insurance to indemnify the railroad for damage. Amish also don't believe in insurance. So that means that there are no crossings on their farms. Driving 5 miles out of your way to get to an existing crossing is a far larger problem if you're on horse than it is if you are driving an internal combustion engine.
One thing that's horribly misleading is when prosecutors say "the likelihood of this match occurring at random is one in a quadrillion" or similar. If there aren't a quadrillion people on the planet, that statement means nothing. Also it's based on a lot of independence assumptions that may or not pan out. The irony is that the answer is out there - with all of the DNA database information that's been compiled by different law enforcement agencies, there is the ability to actually go and test to see whether there are duplicates out there, what the sharing rate is between siblings, twins and parent/children, and so on, so that you can get real measurements instead of probabilistic arguments. If current tests turn out to uniquely identify subjects, the jury should hear "this DNA uniquely identifies this person as its source".
But then as the article points out you also need to turn to the quality control aspect. Identify the potential sources of contamination, quantify those sources through experimental means. Currently agencies do not do blind tests to see what the error rate is in labs. Crime labs should be tested in blind situations to see what their quality rate is. Then you can bring out during the trial "this lab has successfully passed 100 QC tests in the last 2 years and has never failed one" or "this lab failed 2 QC tests out of 100, but the person who failed both has been discharged" or any other information that allows jurors to assess what the error rate is in the other steps in the process. Similarly, success rates are dependent on the size of the sample; if you start from 8 strands of DNA how much does your result degrade when compared with a cheek swab? We just don't have those numbers now, and there's no reason we don't.
DNA is an amazing tool in the crime database. It solves crimes that have not been solved and helps put bad guys behind bars who would have gotten away. But it is not magic or infallible. Quantifying the sources of error and presenting them during trial is the right thing to do.
Put not your trust in money, but put your money in trust.