The late David Galula, whose book "Counterinsurgency Warfare: Theory and Practice" is considered by many to be the bible of counterinsurgency, wrote about his experience more than 50 years ago:
"Demoralization of the enemy's forces is an important task. The most effective way to achieve it is by employing a policy of leniency toward prisoners. They must be well treated and offered the choice of joining the movement or of being set free, even if this means that they will return to the counterinsurgent's side. Despite its setbacks in the early stages, this is the policy that pays the most in the long run."
Galula lived through this firsthand after he was captured in 1947 by Mao Zedong's forces in the Chinese civil war. Rather than being dealt with as a prisoner, he was treated as an "honored guest," as were all Chinese Nationalist soldiers in the camp. They were given two weeks of food and indoctrination, and upon completion of their stay, they had the option to return to their units.
As it turned out, many of the prisoners had been in the camps before -- the knowledge that they would be treated well had led them to view surrender as an acceptable option. Eventually they were viewed with suspicion by their own army, and the Nationalists were forced to set up prisoner of war camps for their own people.
Mao had perfected this technique while fighting the Japanese in World War II; many of the postwar leaders of the Japanese Communist Party were "graduates" of these camps. Mao succeeded in changing the minds of kamikazes -- something for us to think about.
They're right... and many other corruptions of modern American politics could be lessened if incumbents weren't so protected.
We could easily, with today's mapping technology, make contiguous but completely partisan-indifferent districts. As an extreme example, California's 53 districts could each be horizontal stripes, having only latitudinal straight lines as drawn borders, with equal populations. (They'd be wider near Eureka, thin near big cities.) Odd, and tough on campaigners -- but still an improvement over today's noncompetitive districts.
Other algorithms could satisfy 'compactness' as an ease to campaigning, as appropriate -- so long as partisan data is never fed into the algorithm as an input. (So far, the only deployment of advanced demographic technology in practice has been by the parties to assist their gerrymandering.) The Wikipedia article on gerrymandering discusses some possible algorithms.