Casey did a nice post of similar thoughts I’ve had:

OPAC Web Services Should Be Like Amazon Web Services « MaisonBisson.com

I’m going to go a step further and say it should be more than the Amazon Web Service. Since Amazon is a company and has various partners the data you can get through the services are limited though they give you plenty to play with. I’ve ran into walls trying to get some of their data such as New York Times best sellers and other. In general you can build an interface to Amazon but you couldn’t completely recreate Amazon if you wanted to.

I don’t think there should be any such limitation with the OPAC though. The data is most likely yours and I think you should be able to use it how you wish. As such you should have the ability to completely recreate the OPAC if you wish. It’s this kind of data availability that will allow libraries to keep apace with change. Unfortunately the III XML server doesn’t currently have that sort of robust output. It allows you to so some of the simpler things but larger projects require some hacking of the OPAC itself.