From: Scott Schremmer (scotths@cs.washington.edu)
Date: Mon Nov 15 2004 - 01:22:16 PST
Commonly used Internet protocols are designed to function well for
communication between 2 points. They do not work as well when attempting
multicast and other services. Previous solutions to this problem have
either proposed modifications to the IP protocol or applications level
adaptations. Neither is ideal. It is difficult to obtain consensus to
modify IP and application level solutions tend to lead to redundancy
between different applications attempts at implementing a similar
solution.
The paper proposes an "Internet Indirection Infrastructure" which
attempts to find solutions to the problems caused. This approach
attempts to add another layer of abstraction between IP and the
application layer. The key point in this model is the decoupling of
sending and receiving, thus "Indirection". Each communication is assigned
an id number and this is used to send packets rather than the ip address.
Some node in the network matches this id with the ip address or addresses
to which the packet needs to be sent. This allows for mobility as the ip
address could be changed midstream, multicast as it can represent more
that one ip and as well as anycast.
The solution seems quite elegant and is well described and analyzed in the
paper. It does raise some security concerns that need to be studied in
more detail before such a system can be implemented safely.
This archive was generated by hypermail 2.1.6 : Mon Nov 15 2004 - 01:22:16 PST