This is in response to a nodeshell originally titled "Those idiots, they didn't build data redundancy into DNA!"
Actually, they did. The complementary sequence of bases - every A matches with a T, every G matches with a C - allows a double strand of DNA to be regenerated from a single strand. This is what happens when DNA is replicated - the strands are separated and one DNA polymerase complex goes to work on each of the resulting strands, making them double again. More pertinently, if something happens to one of the strands - let's say two thymines next to each other happen to stick together and make a thymine dimer that has to be cut out or excised - the resulting hole can be filled in because DNA polymerase "sees" the two adenines opposite the hole, and knows to fill it in with two thymines.
Diploid organisms (including humans, other vertebrates, and in fact most members of kingdoms other than Monera) have an extra layer of redundancy: if both strands of DNA on one chromosome are damaged,
(say, DNA damage from ultraviolet light), there is still a homologous chromosome with a hopefully identical (or at least similar) sequence. The mechanism of DNA repair by recombination is way too complicated to explain here - in fact, I recall dozing off during those lectures - but basically in the event of a DSB (double-stranded break), the two chromosomes are brought together and do some sort of horizontal mambo that allows DNA polymerase to make new strands for the broken piece of DNA from the sequence on the other chromosome.