Pure bundling better than mixed? or, why doesn't AOL offer standardized dial-up service?

Some dial-up Internet access providers, such as the market leader AOL, require customers to install proprietary connection software to use their service. This is puzzling, because while the software helps certain users, it creates disutility for others (especially expert users and early adopters of Internet service). Why, then, does AOL insist on this connection manager? Why not choose a mixed bundling approach where power users can choose a standardized service that offers access to the Internet and AOL-managed content, and novice users are given a service that integrates access with the connection manager. This paper proposes different possible explanations for why firms might willfully create barriers to entry for customers by pursuing a pure, rather than mixed, bundling strategy where the bundle contains one feature that is negatively valued by a customer segment. Developing economic models for this problem, we propose different hypotheses for explaining the puzzle. We postulate that power users impose exceptionally high cost on AOL's system, and show that the presence of bad customers (whose cost exceeds their valuation) can cripple the profits from the access-only option. Another hypothesis is that even if adding the access-only service improves profits in the short-term, the presence of this service may create a learning effect, where novice users learn to manage without AOL's proprietary software and switch to the cheaper access-only option in future periods.