Benny Yih

I can see the "relative" ranking of features being pretty straighforward, but how fuzzy are the "net present value" numbers ? Are they calculated as ranges with percentage certainty, or with error bars ? Do they come from market surveys of potential customers, on how much more they would pay for widget X with whistle Z ? 8-)

Ted,
Can you elaborate a bit on the deployment choices for JBIS interaction modeling versus a collection of "low-level" web service calls ? Are optimizations such as shared memory messages, instead of HTTPS SOAP, etc. supported ? 8-)