Summary: Programming the Internet

1
Summary Programming the Internet

• Relying on infrastructures/third parties for x
  ...
• Strategy-proof reputation/incentive systems
• Design of protocols/languages/platforms to match
  the context
• Connection with Ad Hoc, Overlays, P2P
• New models of network programmability
• Backward compatible
• Local/billable end-to-end/free (best-effort)
• Connection to measurement/modeling
• Mapping forwarding-time computations onto
  platforms system-level view
• Blue sky/outside the box
• Analog programming for optical, wireless channels
  and interconnects
• Beyond writing programs to breeding them
• Rule-based systems unified support for policies

2
Programming the (Inter)Network

• Do we know how to do it predictably?
• Tussles trust, accounting, QoS, etc.
• Safety complexity comes from composition
• Do we need help from PL/OS communities?
• Do we have a common programming language?
• Compiler techniques, typing, etc.?
• Vendor-independent network processor model

3
Observations Common Themes

• Two very different (but valid!) research contexts
• Existing Internet artifact
• Measurement, overlays, TCP modeling, traffic
  engineering
• Challenge Do not ignore the ugly parts (billing,
  trust, ...)
• Different networks
• Question assumptions, new models
• Challenge Choosing replacement assumptions well
• Policy
• Its all about policy. -- Ed Knightly
• Whom do I trust? Why?
• sociology/economics/anthropology help?
• Specification, management, analysis
• Architectural support
• Importance of values
• Need for validation, reproducibility of results
• Tendency to consider areas done too soon
• Are we doing a good job propagating good values?

4
Observations What We Didnt Hear

• Not much about routing
• Do we fundamentally understand the relationship
  among addressing, topology, state,
  forwarding-time costs, convergence time?
• Real problem is about address assignment
• Suggested additions
• Interactions among routing at different levels