conclusion.tex

% conclusion: sum up contributions, to make sure that the reader got
% all of the key points. Should not be introducing any new material.
\section{Conclusion}
\label{sec:conclusion}
% recap context:
Data formats that define a document object model often rely on a rich
notion of referential context: data structures that identify data
objects, which are themselves used by data objects to refer to each
other.
%
Because referential context is itself is a rich data format, there is
a natural motivation to reuse powerful abstractions defined in the
format for organizing data to define the structure of context itself;
%
done imprecisely, this invites circular definitions and ambiguities,
along with incorrect implementations in practical processors.

% recap contribution:
In this paper, we have performed a detailed study of a critical aspect
of PDF's definition of referential context using cross-reference data.
%
Our work has produced a novel machine-readable definition of the
format's more complicated features for defining cross-reference
streams, namely incremental updates and cross-reference streams,
structured as a well-defined sequence of parsing passes over
increasingly rich semantic values.
%
Our work has resulted in multiple open issues concerning
cross-reference streams submitted to the \emph{International
  Organization for Standardization (ISO)}.
%
We believe that it will be possible to instantiate the structure of
our proposed definition to define and parse other complex formats
that themselves use similar features, namely data offsets and lengths.
%
We believe that such definitions will serve as a foundation for
detecting general problematic structures that may occur in PDF
documents and may occur in such formats as well, including file
cavities.

% point to future work: PDF features:
In the future, we hope to extend our current definition to give a
complete, machine-readable definition of all PDF features that define
referential structure, including cross-reference streams in full
detail, Linearized PDF, and hybrid-reference PDF files.
%
We plan to use such a definition, and its associated parsers, to
% 
\textbf{(1)} deeply inspect security-relevant features of PDF
documents, including freedom from file dead objects and file cavities;
%
\textbf{(2)} rigorously develop implementations that expose document
updates so that a document user is not lead to misinterpret what
document content is attested by a digital signature; and
% 
\textbf{(3)} develop complete validators of higher-level document
structure that manifests in the DOM, and which depends directly on the
integrity of referential structures.