A Postal System Based Digital Network And A Distance Learning System (thesis)

Abstract:

In this thesis, we propose the novel approach of turning storage media
transported by the postal system into a general-purpose and transparent
digital network, extending pervasive, high-bandwidth, and low-cost
connectivity to places such as rural areas in developing countries. We
call such a system the Postmanet. To fully realize its potential,
however, an end user needs better support than being told to burn discs
and toss them into the mail bin. We describe the systems support that
we provide in order to achieve the generality, transparency, efficiency,
and scalability goals of the system. The issues that we address
include: managing ``DVD robots'' that automate mass-processing of DVDs,
application-specific marshaling and unmarshaling of messages, providing
best-effort reliable and secure delivery, simultaneous exploitation of
conventional connectivity, and a mechanism for distributing and updating
application code. Two additional support features are of particular
importance.

The first is a distributed object repository that makes available a
single name space, on which any sites, including those that lack
conventional networking access, can perform read, write, navigation,
search, and other operations. This high-level abstraction makes it
easier to construct distributed Postmanet applications. It also helps
us realize a powerful ``network effect,'' as spontaneous connections are
established among sites that enjoy shared access to a common repository.
The second is scalable routing. Simply leaving end users to directly
swap discs with each other does not scale well, because as many as N2
discs may need to be exchanged at once in an N-node network. We solve
this problem by multiplexing/de-multiplexing data to/from a smaller
number of discs in transit. This can occur multiple times at dedicated
nodes inside the network, or at peer end user nodes. We present routing
topologies that can result in a good balance between simultaneously
minimizing the number of discs involved and the end-to-end postal
latency.

We have built and deployed a real-world application, a rural distance
learning system called the Digital StudyHall, on top of the Postmanet.
It consists of a network of hubs and spokes, where the hubs are urban
centers of excellence, which ``radiate'' content and methodology into
poor villages and slum schools. Our experiences in rural India not only
have provided us insights on the type of the systems support that we
need, but also have allowed us to study mediation-based pedagogy that
has proved promising in extending high-quality education to a needy
population. For more information, please visit:
http://dsh.cs.washington.edu.