\section{Analysis and Conclusions}
\label{analysis}

\subsection{Kitchens}

Only the very beginning of the exploration of a networked kitchen has
begun in this work.  The Hive system proved useful in development of a
flexible, decentralized demonstration networked kitchen.  The
implementation should also provide a good platform on which to extend
and further examine possible applications in the kitchen.

\par

A recurring theme in all attempts to build a ``kitchen of the future''
have been that opinions as to how it should work are strong, varied
and often diametrically opposed.  A componentized system for kitchens,
such as the one built with Hive, will be necessary to experiment with
scenarios to determine which approaches are successful, which are not,
and which depend on the particular user preference.

\subsection{Hive}

Hive has succeeded in allowing a number of complex systems of things
that think to be built.  The kitchen demonstration system alone is the
largest individual Hive application built so far, and it did not run
into any scaling related problems itself.  During the Spring TTT
meeting, a number of other projects within the lab utilized Hive, and
some systems experienced problems related to too many RMI sockets
being produced.  In both the fall and spring demonstrations, many
unanticipated ways of connecting devices together were tried with no
other difficulties.

\par

Further, the Hive system proved to be relatively straightforward for
other users to utilize, despite only minimal documentation and
personalized help being available at the time of the demonstrations.
A number of the other projects, particularly the wearables project
described in Section \ref{wearable} continue to use Hive extensively.

\subsubsection{Semantic Labeling}

The RDF-based semantic labeling approach also proved successful in
application in a small number of systems.  In the kitchen
demonstration, it allowed for flexibility in the use of hardware.
This became especially relevant when the tag reader planned on being
used turned out to be broken, as described in Section
\ref{cianalysis}.

\par

Continuing work using semantic labels will help evaluate whether RDF
is a suitable solution, givens its disadvantages of complexity.  This
may be mitigated or eliminated by the production of RDF tools, either
as part of the Hive project or as general RDF tools.

\subsection{Future Work}

There are many directions in which to continue this research.  Both
the kitchen demonstration and the Hive infrastructure have shown
substantial promise.

\subsubsection{Kitchens}

In the kitchen, an initial area for work is the addition of new
devices to the repertoire of the system.  This will include new devices
similar to those that exist, such as a superior tag reading system, as
well as currently unimplemented devices, such as blenders, stoves,
sinks, and ovens.  An important step in pursuing this is moving the
research into a real kitchen rather than the prototype mock up that
has served so far.  Further, integration of a scheduling system into
the recipe manager, as well as a more sophisticated inventory manager
will provide challenging software projects.

\par

Once an actual kitchen with a standard array of appliances is
available, experiments in usability and stability can begin.  User
interface in the kitchen is of especial importance, as a traditional
screen display is not well suited to the environment.  Various input
and output forms, including projection, speech in and out, use of
wearables, and novel display devices should be explored.

\subsubsection{Hive}

The Hive infrastructure should continue to be developed in a number of
ways.  The current mobility layer is incomplete, and completing this
will allow for a variety of interesting explorations into uses of
mobile code.  Experience with users of Hive has shown that while Hive
is useful and approachable to most users, there is room for a number
of utilities to manage semantic descriptions and configurations, more
documentation for implementors, and a more straightforward method to
install Hive.

\par

Other future developments are likely to include some integration with
a web server to allow ``web application server''-like systems.  A
variety of novel applications of code mobility can be examined,
including a mobile-code-based discovery system and implications to
security.  Partial or full integration with Jini is a further
possibility.  A release outside of the Media Lab is also planned.

\subsection{Conclusion}

Hive has provided the necessary infrastructure to construct a
successful prototype kitchen that can be extended and developed
further.  Additionally, Hive has achieved initial success in making it
substantially easier to build complex networked systems of things that
think.