delay processing of info responses until called for
It would be nice to delay the generation of the info response until it is actually called for. Right now the server initializes and calculates info responses for every data product. This creates a delay and the server cannot be used until the process is done. On my test machine, I've culled the database down to ~100 datasets, and it still takes several minutes to initialize. For the production server I expect this would take 30 minutes to an hour.
This could be done in two steps. First, don't initialize any info responses, but instead check a table of info responses to see if it has been calculated, and then calculate it if necessary. Second, sweeps of HAPI servers will require that all info responses are calculated, so this should be done in the background.