Traditional analysis of data from complex surveys used different software and hardware, and textbooks focused heavily on computational special cases of the Horvitz--Thompson variance formula. This made sense in the past, but now there is very large-scale use of national survey data for research in social and health sciences. With modern computing and modern statistical theory it is possible to hide most of the details from the user, as has been done in Stata and in my 'survey' package for R. The iNZight system for visual data analysis being developed by Chris Wild and co-workers in Auckland will go even further in making the analysis transparent.