It is a common belief that the problem of extracting learners' profiles to be used for delivering custom learning experiences is a closed case. Yet, practical solutions do not completely cope with the complex issue of capturing all the features of users, especially those of heterogeneous learners, who may have special needs or characteristics (such as disabilities), or are exploiting nonstandard equipment to access services (e.g., mobile devices). For example, the standard ACCLIP (Accessibility for Learner Information Package) specification provides means for describing preferences in terms of user accessibility; yet it lacks methods to manage technical characteristics of the exploited device. Conversely, CC/PP (Composite Capabilty/Preferences Profile) profiles are thought to describe device capabilities, but they are unable to report on users' characteristics. With this unsolved problem in view, we present an effective approach to profiling e-learners, which allows the extraction and adaptation of multisource didactical content for customized educational experiences. The idea behind this is to unite the strengths of ACCLIP and CC/PP protocols, while avoiding specification conflicts. Several use cases are described that show the viability of our proposal. (Contains 2 tables and 7 figures.)