diff --git a/Syntax.md b/Syntax.md
index 600935d..90d8b16 100644
--- a/Syntax.md
+++ b/Syntax.md
@@ -121,11 +121,11 @@ To produce the familiar bracketed references to other specifications, e.g.
 
 > There are a number of dynamic selectors that can be used with HTML. This section defines when these selectors match HTML elements. [[SELECTORS]](https://html.spec.whatwg.org/#refsSELECTORS) [[CSSUI]](https://html.spec.whatwg.org/#refsCSSUI)
 
-you can use the `<ref>` void element:
+you can use the (nonstandard) `<ref>` element:
 
 ```html
 <p>There are a number of dynamic selectors that can be used with HTML. This section defines when
-these selectors match HTML elements. <ref spec=SELECTORS> <ref spec=CSSUI></p>
+these selectors match HTML elements. <ref>SELECTORS</ref> <ref>CSSUI</ref></p>
 ```
 
 These match against a bibliography, which is a manually-maintained and sorted `<dl>` at the end of the source file, identified by having the ID "`ref-list`". Its entries look like the following:
diff --git a/src/html/htmlparser.pas b/src/html/htmlparser.pas
index c7bea3f..08460ca 100644
--- a/src/html/htmlparser.pas
+++ b/src/html/htmlparser.pas
@@ -219,7 +219,6 @@       TToken = record // this is an expensive type, as it is basically all token
      const
       Marker = nil;
      var
-      FProprietaryVoids: specialize PlasticArray <TCanonicalString, TCanonicalString>;
       FInputStream: TInputStream;
       {$IFDEF PARSEERROR} FOnParseError: TParseErrorHandler; {$ENDIF}
       FTokeniserState: TTokeniserState;
@@ -314,7 +313,6 @@       TToken = record // this is an expensive type, as it is basically all token
       procedure SpoonFeed(const Data: UTF8String); // call this any number of times until all characters have been provided
       {$ENDIF}
       procedure SpoonFeed(const Data: Pointer; const Length: QWord); // call this any number of times until all characters have been provided
-      procedure RegisterProperietaryVoidElements(const TagNames: array of TCanonicalString);
       function Parse(): TDocument; // then call this
       // XXX need a fragment parsing mode (if we support fragment parsing, set FFragmentParsingMode to true)
       {$IFDEF PARSEERROR} property OnParseError: TParseErrorHandler read FOnParseError write FOnParseError; {$ENDIF}
@@ -1136,15 +1134,6 @@ procedure THTMLParser.SpoonFeed(const Data: Pointer; const Length: QWord);
    FInputStream.PushData(Data, Length);
 end;
 
-procedure THTMLParser.RegisterProperietaryVoidElements(const TagNames: array of TCanonicalString);
-var
-   Name: TCanonicalString;
-begin
-   {$IFOPT C+} Assert(not FInputStream.WasStarted); {$ENDIF}
-   for Name in TagNames do
-      FProprietaryVoids.Push(Name);
-end;
-
 function THTMLParser.Parse(): TDocument;
 var
    OldKind: TTokenKind;
@@ -1936,7 +1925,7 @@ procedure THTMLParser.Tokenise();
          end;
          BogusComment();
       end;
-      
+
       procedure TryForCDATASection(); inline;
       begin
          // seen [
@@ -2125,7 +2114,7 @@ procedure THTMLParser.Tokenise();
       CommentSize := 0;
       repeat
          case (FInputStream.CurrentCharacter.Value) of
-            $003E, kEOF: break; 
+            $003E, kEOF: break;
             $0000: Inc(CommentSize, FFFD.Length);
             else Inc(CommentSize, FInputStream.CurrentCharacterLength);
          end;
@@ -2562,7 +2551,7 @@ procedure THTMLParser.Tokenise();
                         {$IFDEF PARSEERROR} ParseError('unexpected U+0000 in attribute name'); {$ENDIF}
                         FCurrentToken.CurrentAttributeName.Append($FFFD);
                      end;
-                  $0022, $0027, $003C: 
+                  $0022, $0027, $003C:
                      begin
                         {$IFDEF PARSEERROR} ParseError('invalid character in attribute name'); {$ENDIF}
                         FCurrentToken.CurrentAttributeName.Append(FInputStream.CurrentCharacter);
@@ -5907,7 +5896,7 @@ procedure THTMLParser.TheInBodyInsertionMode(var Token: TToken);
          if (Token.TagName = eBody) then
          begin
             {$IFDEF PARSEERROR} ParseError('unexpected body start tag'); {$ENDIF}
-            if ((FStackOfOpenElements.Length < 2) or 
+            if ((FStackOfOpenElements.Length < 2) or
                 (not FStackOfOpenElements[1].IsIdentity(nsHTML, eBody)) or
                 (StackOfOpenElementsHas(nsHTML, eTemplate))) then
                exit; // ignore the token
@@ -5930,7 +5919,7 @@ procedure THTMLParser.TheInBodyInsertionMode(var Token: TToken);
          if (Token.TagName = eFrameset) then
          begin
             {$IFDEF PARSEERROR} ParseError('unexpected body frameset start tag'); {$ENDIF}
-            if ((FStackOfOpenElements.Length < 2) or 
+            if ((FStackOfOpenElements.Length < 2) or
                 (not FStackOfOpenElements[1].IsIdentity(nsHTML, eBody))) then
                exit; // ignore the token
             if (not FFramesetOkFlag) then
@@ -6298,11 +6287,6 @@ procedure THTMLParser.TheInBodyInsertionMode(var Token: TToken);
             // any other start tag
             ReconstructTheActiveFormattingElements();
             InsertAnHTMLElementFor(Token);
-            if (FProprietaryVoids.Contains(Token.TagName)) then
-            begin
-               FStackOfOpenElements.Pop();
-               {$IFDEF PARSEERROR} Token.AcknowledgeSelfClosingFlag(); {$ENDIF}
-            end;
          end;
       tkEndTag:
          // in this section things are hoisted also
@@ -6381,7 +6365,7 @@ procedure THTMLParser.TheInBodyInsertionMode(var Token: TToken);
          if ((Token.TagName = eDiv) or
              (Token.TagName = ePre) or
              (Token.TagName = eOL) or
-             (Token.TagName = eDL)) then 
+             (Token.TagName = eDL)) then
          begin
             if (not StackOfOpenElementsHasInScope(Token.TagName)) then
             begin
@@ -6467,7 +6451,7 @@ procedure THTMLParser.TheInBodyInsertionMode(var Token: TToken);
              //(Token.TagName = ePre) or // hoisted
              (Token.TagName = eSection) or
              (Token.TagName = eSummary) or
-             (Token.TagName = eUL)) then 
+             (Token.TagName = eUL)) then
          begin
             if (not StackOfOpenElementsHasInScope(Token.TagName)) then
             begin
@@ -6815,7 +6799,7 @@ procedure THTMLParser.TheInTableInsertionMode(var Token: TToken);
             exit;
          end
          else
-         if ((Token.TagName = eBody) or 
+         if ((Token.TagName = eBody) or
              (Token.TagName = eCaption) or
              (Token.TagName = eCol) or
              (Token.TagName = eColGroup) or
@@ -7869,7 +7853,7 @@ procedure THTMLParser.TheAfterFramesetInsertionMode(var Token: TToken);
             end;
          end;
       tkExtraSpaceCharacter: InsertCharacters(Token.ExtraChars);
-      {$IFDEF PARSEERROR} tkExtraCharacters: ParseError('unexpected character token after frameset', Length(Token.ExtraChars)); {$ENDIF} // $R- 
+      {$IFDEF PARSEERROR} tkExtraCharacters: ParseError('unexpected character token after frameset', Length(Token.ExtraChars)); {$ENDIF} // $R-
       tkComment: InsertAComment(Token);
       {$IFDEF PARSEERROR} tkDOCTYPE: ParseError('unexpected DOCTYPE'); {$ENDIF}
       tkStartTag:
@@ -8131,49 +8115,49 @@ procedure THTMLParser.TheRulesForParsingTokensInForeignContent(var Token: TToken
       tkComment: InsertAComment(Token); // http://bugs.freepascal.org/view.php?id=26403
       {$IFDEF PARSEERROR} tkDOCTYPE: ParseError('unexpected DOCTYPE'); {$ENDIF}
       tkStartTag:
-         if ((Token.TagName = eB) or 
-             (Token.TagName = eBig) or 
-             (Token.TagName = eBlockQuote) or 
-             (Token.TagName = eBody) or 
-             (Token.TagName = eBr) or 
-             (Token.TagName = eCenter) or 
-             (Token.TagName = eCode) or 
-             (Token.TagName = eDD) or 
-             (Token.TagName = eDiv) or 
-             (Token.TagName = eDL) or 
-             (Token.TagName = eDT) or 
-             (Token.TagName = eEm) or 
-             (Token.TagName = eEmbed) or 
-             (Token.TagName = eH1) or 
-             (Token.TagName = eH2) or 
-             (Token.TagName = eH3) or 
-             (Token.TagName = eH4) or 
-             (Token.TagName = eH5) or 
-             (Token.TagName = eH6) or 
-             (Token.TagName = eHead) or 
-             (Token.TagName = eHR) or 
-             (Token.TagName = eI) or 
-             (Token.TagName = eImg) or 
-             (Token.TagName = eLI) or 
-             (Token.TagName = eListing) or 
-             (Token.TagName = eMenu) or 
-             (Token.TagName = eMeta) or 
-             (Token.TagName = eNoBr) or 
-             (Token.TagName = eOL) or 
-             (Token.TagName = eP) or 
-             (Token.TagName = ePre) or 
-             (Token.TagName = eRuby) or 
-             (Token.TagName = eS) or 
-             (Token.TagName = eSmall) or 
-             (Token.TagName = eSpan) or 
-             (Token.TagName = eStrong) or 
-             (Token.TagName = eStrike) or 
-             (Token.TagName = eSub) or 
-             (Token.TagName = eSup) or 
-             (Token.TagName = eTable) or 
-             (Token.TagName = eTT) or 
-             (Token.TagName = eU) or 
-             (Token.TagName = eUL) or 
+         if ((Token.TagName = eB) or
+             (Token.TagName = eBig) or
+             (Token.TagName = eBlockQuote) or
+             (Token.TagName = eBody) or
+             (Token.TagName = eBr) or
+             (Token.TagName = eCenter) or
+             (Token.TagName = eCode) or
+             (Token.TagName = eDD) or
+             (Token.TagName = eDiv) or
+             (Token.TagName = eDL) or
+             (Token.TagName = eDT) or
+             (Token.TagName = eEm) or
+             (Token.TagName = eEmbed) or
+             (Token.TagName = eH1) or
+             (Token.TagName = eH2) or
+             (Token.TagName = eH3) or
+             (Token.TagName = eH4) or
+             (Token.TagName = eH5) or
+             (Token.TagName = eH6) or
+             (Token.TagName = eHead) or
+             (Token.TagName = eHR) or
+             (Token.TagName = eI) or
+             (Token.TagName = eImg) or
+             (Token.TagName = eLI) or
+             (Token.TagName = eListing) or
+             (Token.TagName = eMenu) or
+             (Token.TagName = eMeta) or
+             (Token.TagName = eNoBr) or
+             (Token.TagName = eOL) or
+             (Token.TagName = eP) or
+             (Token.TagName = ePre) or
+             (Token.TagName = eRuby) or
+             (Token.TagName = eS) or
+             (Token.TagName = eSmall) or
+             (Token.TagName = eSpan) or
+             (Token.TagName = eStrong) or
+             (Token.TagName = eStrike) or
+             (Token.TagName = eSub) or
+             (Token.TagName = eSup) or
+             (Token.TagName = eTable) or
+             (Token.TagName = eTT) or
+             (Token.TagName = eU) or
+             (Token.TagName = eUL) or
              (Token.TagName = eVar) or
              ((Token.TagName = eFont) and (Token.HasAttributes(['color', 'face', 'size'])))) then
          begin
diff --git a/src/wattsi.pas b/src/wattsi.pas
index 7cb5694..e7dad94 100644
--- a/src/wattsi.pas
+++ b/src/wattsi.pas
@@ -1579,8 +1579,7 @@    TCrossReferences = record
          else
          if (Element.IsIdentity(nsHTML, eRef)) then
          begin
-            ExtractedData := Element.GetAttribute('spec');
-            ReferenceName := ExtractedData.AsString;
+            ReferenceName := Element.TextContent.AsString;
             New(ListNode);
             ListNode^.Value := Element;
             ListNode^.Next := References[ReferenceName];
@@ -1589,12 +1588,11 @@    TCrossReferences = record
             NewLink := ConstructHTMLElement(eA);
             Scratch := Default(Rope);
             Scratch.Append('#refs');
-            Scratch.AppendDestructively(ExtractedData); // $R-
+            Scratch.Append(ReferenceName);
             NewLink.SetAttributeDestructively('href', Scratch);
-            ExtractedData := Element.GetAttribute('spec');
             Scratch := Default(Rope);
             Scratch.Append('[');
-            Scratch.AppendDestructively(ExtractedData); // $R-
+            Scratch.Append(ReferenceName);
             Scratch.Append(']');
             NewLink.AppendChild(TText.CreateDestructively(Scratch));
             (Node.ParentNode as TElement).ReplaceChild(NewLink, Node);
@@ -2961,7 +2959,6 @@ function Main(): Boolean;
    Source := ReadFile(SourceFile);
    try
       Parser := THTMLParser.Create();
-      Parser.RegisterProperietaryVoidElements([eRef]);
       try
          try
             Parser.SpoonFeed(Source.Start, Source.Length);