This publication analyses the requirements that user modeling servers must meet to be acceptable both from a multi-disciplinary scientific perspective and from a (commercial) deployment perspective. The focus is also on the design and implementation of servers that meet these requirements and on the verification of the compliance with the core performance and scalability of small and medium-sized, real world environments. Software systems that adapt their services to characteristics of individual users have already proven to be more effective and useful than non-adaptive systems. For more personalized features, the user-adaptive system relies on models of user characteristics. Effective acquisition and management of these models is carried out by specific user modeling components. An important strand of research is dedicated to the development of the so-called user modeling shell system’. The decision as to what the services and functionalities these shell systems should have, is mostly based on intuition and experience gained from studying literature concerning user-adaptive applications. More recently, the trend towards personalization on the World Wide Web led to the development of several commercial user modeling servers. The difficulty is that features that will be very important to these systems show a sharp contract with those regarded as being important for a user modeling shell systems, and vice versa.