Just some added info for people unfamiliar with how most of these visual approaches are done:
There is the General aviation Visual Approach, were the center controller clears you direct to the airport, you see the field, inform ATC, he cancels ATC services, you find your own way to the airport, looking and avoiding other traffic on your own. That is not what SFO approach wanted from DLH.
For 121, scheduled passenger flights to big airports, everyone is vectored/altitude/speed controlled by ATC, until on a reasonable heading to intercept the straight in approach you will use to back-up the visual. You call the field in sight, and intercept the normal LOC/CRS followed by the GS/GP. The reason for the visual is things like reduced spacing,or less workload for controllers. In SFO it is because the parallel runways are quite close, so in order to use both simultaneously, without having to use PRM/SOIA approaches (more dangerous in my opinion), they use "visual" approaches. I firmly believe ATC should have worked with DLH better. But I am equally convinced DLH should have made an exception for the way these approaches are conducted at major US airports. When I flew in the EU, in night VMC we were not allowed to go below the MSA/MDA/MAA/MOCA/MEA/MVA. That makes sense. The DLH SOP does not.