diff --git a/LineTAP.tex b/LineTAP.tex
index 547f9d9..3734305 100644
--- a/LineTAP.tex
+++ b/LineTAP.tex
@@ -284,7 +284,7 @@ \subsection{Non-Use Cases}
 \rowsep
 \texttt{xsams\_uri} \hfil\break\ucd{meta.ref} & text & \raggedright A URI for a full XSAMS description of this line.\tabularnewline
 \rowsep
-\texttt{line\_reference} \hfil\break\ucd{meta.ref} & text & \raggedright Reference to the source of the line data; this could be a bibcode, a DOI, or a plain URI.\tabularnewline
+\texttt{line\_reference} \hfil\break\ucd{meta.ref} & \textbf{text} & \raggedright Reference to the source of the line data; this could be a bibcode, a DOI, or a plain URI.\tabularnewline
 
 % /GENERATED
 \sptablerule
@@ -391,7 +391,9 @@ \section{Spectral Lines Table}\label{sect:quantities}
 return exactly one transition, and it MUST be in XSAMS version 1.
 
 \item \texttt{line\_reference} Information about the source of the line data,
-like an URI, DOI or bibcode.
+like an URI, DOI or bibcode.  This is a mandatory column.  Users should
+have a guaranteed way of finding out where a piece of information came
+from and where additional information is available.
 
 \end{itemize}
 
diff --git a/make-columns-table.py b/make-columns-table.py
index 01c9339..319617e 100644
--- a/make-columns-table.py
+++ b/make-columns-table.py
@@ -9,7 +9,7 @@
 
 import pyvo
 
-NON_NULL_COLUMNS = {'title', 'vacuum_wavelength'}
+NON_NULL_COLUMNS = {'title', 'vacuum_wavelength', 'line_reference'}
 TYPE_MAP = {
 	("char", "*"): "text",
 	("int", ""): "integer",
@@ -17,7 +17,7 @@
 
 
 def e(tx):
-	"""returns tx with TeX's standard active (and other magic) characters 
+	"""returns tx with TeX's standard active (and other magic) characters
 	escaped.
 	"""
 	return tx.replace("\\", "$\\backslash$"
@@ -49,7 +49,7 @@ def main():
 	for row in svc.run_sync("""
 			select column_name, description, unit, ucd, datatype, arraysize
 			from tap_schema.columns
-			where 
+			where
   			table_name='casa_lines.line_tap'
   			and std=1
 			order by column_index"""):
@@ -61,8 +61,8 @@ def main():
 		
 		parts.append("&")
 		parts.append(get_type(
-			row["datatype"], 
-			row["arraysize"], 
+			row["datatype"],
+			row["arraysize"],
 			row["column_name"] in NON_NULL_COLUMNS))
 
 		parts.append("&")