Bringing the past to the present for a better tomorrow.

\section{Introduction}

\section{Related Work}

\subsection{Snapshotting}

Short description, possibly a single diagram

List of example systems, preferably in a bulleted list with short
descriptions, though a simple list will suffice

Emphasize non-continuousness: can't really compare representations
with PersiFS (except that they're usually not too bad about
space-efficiency)

\subsection{Logging}

Short description, possibly a single diagram.  ``Logging systems still
require some form of periodic snapshots to prevent having the replay
the log from the beginning.''

Slip in somehow that both of these require maintaining a separate copy
of the current version.

List of example systems

\subsection{Interface}

PersiFS doesn't need special tools (transparent file system access).
and it has a timestamp-based interface, which I assert (but don't
really want to argue in this poster) is the right thing to do.

Contrast to systems that don't do this.

Maybe a good way to handle related work is with a feature table.

\section{Advantages of PersiFS}

PersiFS is continuous, space-efficient, time-efficient, the current
version is maintained in the same data structures as past versions,
etc.

Where to put asymptotic bounds?  They should be somewhere primary, but
might be best after the structures are described (OTOH, we could say
``It achieves this, and here's how...'')

\section{PersiFS Structures}

Maybe a blurb about how PersiFS is composed.  I think we have a
diagram of this. (Add the superblob to the diagram from the end of the
897 report?)

Might be better to name the subsections after the problems they're
solving (time efficiency and space efficiency?) instead or in addition
to the structures that achieve them.

\subsection{The Index Tree} or whatever we called it...

Let's not call it the arborescent metadata map. I like it, but...
Maybe 'persistent metadata tree' in order to hammer home the
persistence idea? (We didn't give it a name in the abstract)

Explain/define persistent data structures

Define the interface: inumber -> metadata map, modify current
revision, lookup on current or previous. Maybe mention explicit
commit, or maybe not?

Be sure to explain what the cost is logarithmic in. Compare against
normal tree, log.

\subsection{The Superblob}

\section{Future Work}

\begin{itemize}
\item \textbf{Retention policy} -- Foo
\end{itemize}

Or maybe a describe environment.  Probably best to avoid a plain text
block, unless perhaps the key phrases are bolded.

\section{References}