As I was reading the paper I discussed yesterday, I realized there was a part I didn't fully understand and I needed to research more. I received some great comments on that post that pointed me in the right direction.
A gene delivery vector is an engineered virus modified so that it contains the genes needed for therapy. Once inside the cell, the genetic material needs to reach the nucleus where it has to recruit a complementary DNA strand in order for the gene to be expressed.
Conceptually, it seems easy enough. In practice, every step of the way has its hurdles and of all the vectors you inject into the host, only a fraction turns into expressed genes. With adeno-associated virus (AAV) the major bottleneck is the de novo synthesis of the DNA: not all the single-stranded DNA delivered to the nucleus is successfully converted into double-stranded DNA, thus hindering the efficiency of the vector.
Luckily, there's a few alternatives. Suppose you have two viruses, and each carries complementary DNA. Assuming they both reach the nucleus, the two DNA strands will find each other (no need to recruit a complementary strand from the existing chromatin), and voila' -- you have a double-stranded DNA. Now, this in general wouldn't be possible with just any virus, because they tend to have a preference for which strand they carry. But with AAV we're in luck because it packages either strand with equal efficiency.
This approach is also prone to issues. For example, it's hard to predict whether or not the two complementary strands, once inside the nucleus, will find one another. The likelihood increases with the dose, but that also increases the chances of recombination.
What about packaging both strands inside the same vector? Turns out it's possible, even though you lose in capacity (you can't package as much DNA inside the virus, approximately half of what you could achieve with the previous AAV). As McCarty explains in :
"This can be achieved by taking advantage of the tendency to produce dimeric inverted repeat genomes during the AAV replication cycle. If these dimers are small enough, they can be packaged in the same manner as conventional AAV genomes, and the two halves of the ssDNA molecule can fold and base pair to form a dsDNA molecule of half the length. Although this further restricts the transgene carrying capacity of an already small viral vector, it offers a substantial premium in the efficiency, and speed of onset, of transgene expression because dsDNA conversion is independent of host-cell DNA synthesis and vector concentration."
The above figure shows the steps through which this is achieved. Technical details aside, this new mechanism exploits the virus's ability to naturally form short complementary strands. The "shortness" diminishes the capacity, but if you can get away with delivering short genes, then you can efficiently bypass the de novo synthesis bottleneck and greatly increase the efficiency of the vector. These are called self-complementary vectors, scAAV in the case of adeno-associated virus.
There are still hurdles to circumvent. You still face the potential barrier posed by humoral immunity, the fact that the immune system might recognize the virus and destroy it before it can reach its destination. In  McCarty reviews several applications of scAAV, including cell lines where studies have been successful, and others that haven't. But the paper I discussed yesterday was certainly a great step forward and a success story in the use of scAAV vectors.
 McCarty, D. (2008). Self-complementary AAV Vectors; Advances and Applications Molecular Therapy, 16 (10), 1648-1656 DOI: 10.1038/mt.2008.171