+ & & l \\
+ & | & S~\verb+\sub+~S\\
+ & | & S~\verb+\sup+~S\\
+ & | & S~\verb+\below+~S\\
+ & | & S~\verb+\atop+~S\\
+ & | & S~\verb+\over+~S\\
+ & | & S~\verb+\atop+~S\\
+ & | & \verb+\frac+~S~S \\
+ & | & \verb+\sqrt+~S \\
+ & | & \verb+\root+~S~\verb+\of+~S \\
+ & | & \verb+(+~P~\verb+)+ \\
+ & | & \verb+hbox (+~P~\verb+)+ \\
+ & | & \verb+vbox (+~P~\verb+)+ \\
+ & | & \verb+hvbox (+~P~\verb+)+ \\
+ & | & \verb+hovbox (+~P~\verb+)+ \\
+ & | & \verb+break+ \\
+ & | & \verb+list0+~S~[\verb+sep+~l] \\
+ & | & \verb+list1+~S~[\verb+sep+~l] \\
+ & | & \verb+opt+~S \\
+ & | & [\verb+term+]~x \\
+ & | & \verb+number+~x \\
+ & | & \verb+ident+~x \\
+\end{array}
+\]
+\hrule
+\end{table}
+
+Rationale: while the layout schemata can occur in the concrete syntax
+used by user, the box schemata and the magic patterns can only occur
+when defining the notation. This is why the layout schemata are
+``escaped'' with a backslash, so that they cannot be confused with
+plain identifiers, wherease the others are not. Alternatively, they
+could be defined as keywords, but this would prevent their names to be
+used in different contexts.
+
+\begin{table}
+\caption{\label{tab:l1a} Abstract syntax of level 1 terms and patterns.\strut}
+\hrule
+\[
+\begin{array}{@{}ll@{}}
+\begin{array}[t]{rcll}
+ T & ::= & & \mbox{(\bf terms)} \\
+ & & L_\kappa[T_1,\dots,T_n] & \mbox{(layout)} \\
+ & | & B_\kappa^{ab}[T_1\cdots T_n] & \mbox{(box)} \\
+ & | & \BREAK & \mbox{(breakpoint)} \\
+ & | & \FENCED{T_1\cdots T_n} & \mbox{(fenced)} \\
+ & | & l & \mbox{(literal)} \\[2ex]
+ P & ::= & & \mbox{(\bf patterns)} \\
+ & & L_\kappa[P_1,\dots,P_n] & \mbox{(layout)} \\
+ & | & B_\kappa^{ab}[P_1\cdots P_n] & \mbox{(box)} \\