Development of Intelligent Environments have been so far mostly ad-hoc and here we investigate one fundamental bottleneck which needs to be addressed to facilitate more effective developments in the future: handling of user preferences in a multiple user environment. This paper analyzes some cases which combine different approaches to manage users preferences combining services at higher and lower levels with user-led and environment-led approaches. We assess some practical pros and cons in each of these combinations as well as some more fundamental building blocks which developers need to reflect on from a scientific point of view before embarking on the engineering of these systems.