From: Rosalia Tungaraza (rltungar@u.washington.edu)
Date: Sun Nov 14 2004 - 23:29:35 PST
This paper is about i3, which is an overlay-based Internet indirection
service. The major goal of i3 is to abstract the notion of sending IP
packets from a sender to a receiver through the use of the
overlay-network, removing the IP tags i.e. packets associated with an
identifier that is sent to specific servers. The servers in turn know how
to direct the packet to either its destination or to another server that
knows the destination of that packet. By using this infrastructure the
authors claim that services such as multicasting, anycasting, and mobility
would be better served compared to the current Internet's infrastructure.
One of the successes of this paper is that to some degree it does provide
a better way (from end-users) to carry out multicasting compared to how it
is done now. In i3 they only need to register a trigger with the same
identifier among all members of a multicast. This enables them to receive
packets targeted for that identifier. By the same token, i3 apparently
doesn't differentiate between multicast and unicast in sending or
receiving as IP multicast does. Thus, i3 offers more flexibility and seems
more practical than IP multicast.
One point I would have liked them to expand is how feasible an
implementation of i3 on the Internet is. In other words, will the
additional network physical equipments and software required to mount i3
on top of the existing network be feasible. Are there any aspects of the
current Internet that needs to evolve/migrate in order to be compatible
with the i3 architecture?
The authors propose a number of items for future work in this field. Among
those, they suggest to explore the types/ranges of services and
applications that can be synthesized from the fixed abstraction provided
by i3. They also propose to implement i3 on top of other lookup protocols
such as CAN, Pastry, and Tapestry.
This archive was generated by hypermail 2.1.6 : Sun Nov 14 2004 - 23:29:52 PST