When faced with the term "mobile agent", most software developers imagine, at best, maintenance, visualization, or debugging problems that would make mobile agents difficult to use in any production application. At worst, they imagine security problems that would make mobile agents only one step short of viruses and worms. Even those software developers who recognize these two visions as exaggerated often dismiss mobile agents as a technology that does not provide any benefits over their normal distributed-computing practices. These developer biases arise from reasons both simple — early unrealized hype and an unfortunate name — and profound — legitimate concerns about security, efficiency, and programmer convenience. To overcome the biases, the mobile-agent community must view their current systems only as excellent vehicles for exploring the hard research problems, and must make an intense effort to move individual mobile-code ideas and techniques into simpler, but initially more palatable, mobile-code systems.