What kind of service do end-users need to deal with spam email, and who should implement that service (according to the end-to-end argument)? Are there opportunities for substantial optimizations by pushing functionality into the network?
When you describe the service, please specify the interface between the end-user and the service. For example, does the service deliver "cleaned email" to the end-user (it determines which email is spam and discards it, without notifying the end-user about it)? Does the end-user specify to the service whom (s)he wants to receive email from, i.e., a "whitelist"?
When you describe the implementation of the service, please specify whether it will involve functionality at the end host (e.g., the email client), the email servers, the routers, etc. Where appropriate, argue why you chose to deploy the specific functionality at the specific point. Assume that you can change the software/network stack at the end-hosts, as well as make changes to network equipment.