Comments - 1 Motivating examples More control over routing decisions Bootstrapping networks: Zero device-specific config Joint control of data plane mechanisms Detecting/preventing end-system created security holes Dissemination plane – any two nodes with a physically connected path should be able to communicate Should this be a goal to shoot for? Does it require research or just engineering?
Comments - 2 Does the e2e principle have anything to say about control/management? Minimize what 4D does – boot strap the network and then use existing mechanisms/management systems? RSVP With a broader view of both data (distributed) and VoIP (centralized), It's daunting to tackle the full complexity that arises
Research Problems Is it important to have a way to always talk to all nodes in the network so long as there is a physical path between the nodes? What is the best way to achieve this? How deep should discovery be pushed into the end-systems? Is there a general abstraction that can express the kinds of configurations seen today and the likely into the future? How should decision logic be structured?
Concrete Steps Establish classic problems to use as benchmarks Integrating a new router into a network using OSPF as an IGP Establish measures for success How to compare two different architectures? Create a problem-formulation and call-to-arms paper