diff --git a/raterspec.cls b/raterspec.cls
index 5f58e2d..94be281 100644
--- a/raterspec.cls
+++ b/raterspec.cls
@@ -121,6 +121,10 @@
   {}
   {}
 
+\@raterspec@mkhook{dataclass}{data classifications}
+  {}
+  {}
+
 
 
 %%
diff --git a/raterspec/parts.tex b/raterspec/parts.tex
new file mode 100644
index 0000000..5a9c35c
--- /dev/null
+++ b/raterspec/parts.tex
@@ -0,0 +1,35 @@
+% raterspec common structure
+%%
+
+%% note: \pnumoff and \pnumon are temporary kluges (...hopefully)
+
+
+\pnumoff\part{General System}\pnumon
+
+\chapter{Host Environment}
+  \input{raterspec/sec/hostenv}
+
+\chapter{Input Data}
+  \index{input data|(}
+    \input{raterspec/sec/input}
+
+    % implementation-specific parameter types
+    \section{Program-Specific Parameter Types}
+    \sectiondept{it/uw}
+
+    \inputtypesout
+  \index{input data|)}
+
+\chapter{Input Parameters}
+  \index{input data!parameters|(}
+  \index{parameter|(}
+    \input{raterspec/sec/params}
+    \inputparamsout
+
+  \index{input data!parameters|)}
+  \index{parameter|)}
+
+
+\pnumoff\part{Data Classification}\pnumon
+  \input{raterspec/sec/dataclass}
+
diff --git a/raterspec/sec/dataclass.tex b/raterspec/sec/dataclass.tex
new file mode 100644
index 0000000..f380874
--- /dev/null
+++ b/raterspec/sec/dataclass.tex
@@ -0,0 +1,69 @@
+% Data classifications
+
+\label{s:dataclass}
+\index{classification|(}
+
+\chapter{Common Classifications}
+
+\label{s:dataclass:common}
+
+% applies until first subsection
+\sectiondept{it}
+
+\incomplete
+A classification---also known as a ``class'', but that term is avoided in this
+specification due to ambiguities---is a set of rules performed on the ^[input
+data] yielding a boolean result of the same depth as the largest depth of the
+applicable ^[input data]. \todo{Include formal documentation from DSL manual.}
+
+\begin{ex}
+  Consider a classification~$c$ that has the following rules:
+  \begin{itemize}
+    \item $5\leq\alpha\leq10$
+    \item $\beta>0$
+  \end{itemize}
+  Given various sets of ^[input data], we would expect the results shown in
+  \fref{class-ex}.
+\end{ex}
+
+\begin{figure}
+  \center
+  \begin{tabular}{cc|c}
+    $\alpha$ & $\beta$ & Result \\
+    \hline
+    1 & 2 & $\bot$ \\
+    5 & 2 & $\top$ \\
+
+    \set{4,5} & \set{10,10} & \set{\bot,\top} \\
+    \set{\set{4,5},\set{5,6}} & \set{10,0}
+      & \set{\set{\bot,\top},\set{\bot,\bot}} \\
+  \end{tabular}
+
+  \caption{Results of various input data some classification}
+  \label{f:class-ex}
+\end{figure}
+
+A classification may treat the result of another classification as ^[input
+data].
+
+A classification may treat the result of a~^[calculation] (see \sref{premcalc}) as
+^[input data].
+
+All classifications in this section \shall apply to the ^[input data] as defined
+in~\sref{indata}.
+
+Each classification \shall have an associated character string
+\dfn{classification!description} which \shall be made accessible to the caller
+as ^[output data] in an \unspecified\ manner.
+
+The point at which classifications are calculated is \unspecified; an
+implementation \may choose to defer any specific classification until such a
+time that it is needed by a calculation. Furthermore, an implementation \may
+choose not to perform a given classification at all, so long as it is determined
+that such a classification does not apply to any ^[input data].
+
+
+%% user content
+\dataclassout
+
+\index{classification|)}