diff --git a/src/document/dom/renderer.c b/src/document/dom/renderer.c
index 882ac7ef..93934917 100644
--- a/src/document/dom/renderer.c
+++ b/src/document/dom/renderer.c
@@ -507,8 +507,8 @@ render_dom_element_end_source(struct dom_stack *stack, struct dom_node *node, vo
 	struct dom_stack_state *state = get_dom_stack_top(stack);
 	struct sgml_parser_state *pstate = get_dom_stack_state_data(stack->contexts[0], state);
 	struct dom_scanner_token *token = &pstate->end_token;
-	unsigned char *string = token->string;
-	int length = token->length;
+	unsigned char *string = token->string.string;
+	int length = token->string.length;
 
 	assert(node && renderer && renderer->document);
 
diff --git a/src/dom/css/scanner.c b/src/dom/css/scanner.c
index 3950814d..d745b553 100644
--- a/src/dom/css/scanner.c
+++ b/src/dom/css/scanner.c
@@ -130,7 +130,7 @@ scan_css_token(struct dom_scanner *scanner, struct dom_scanner_token *token)
 	int real_length = -1;
 
 	assert(first_char);
-	token->string = string++;
+	token->string.string = string++;
 
 	if (is_css_char_token(first_char)) {
 		type = first_char;
@@ -174,7 +174,7 @@ scan_css_token(struct dom_scanner *scanner, struct dom_scanner_token *token)
 			/* Make sure that we have an ending ')' */
 			skip_css(scanner, function_end, ')');
 			if (*function_end == ')') {
-				type = map_dom_scanner_string(scanner, token->string,
+				type = map_dom_scanner_string(scanner, token->string.string,
 						string, CSS_TOKEN_FUNCTION);
 
 				/* If it is not a known function just skip the
@@ -202,7 +202,7 @@ scan_css_token(struct dom_scanner *scanner, struct dom_scanner_token *token)
 					if (isquote(*from)) from++;
 					if (isquote(*to)) to--;
 
-					token->string = from;
+					token->string.string = from;
 					real_length = to - from + 1;
 					assert(real_length >= 0);
 					string = function_end;
@@ -233,7 +233,7 @@ scan_css_token(struct dom_scanner *scanner, struct dom_scanner_token *token)
 			/* Check that the hexdigit sequence is either 3 or 6
 			 * chars and it isn't just start of some non-hex ident
 			 * string. */
-			hexdigits = string - token->string - 1;
+			hexdigits = string - token->string.string - 1;
 			if ((hexdigits == 3 || hexdigits == 6)
 			    && !is_css_ident(*string)) {
 				type = CSS_TOKEN_HEX_COLOR;
@@ -299,8 +299,8 @@ scan_css_token(struct dom_scanner *scanner, struct dom_scanner_token *token)
 
 		if (string_end) {
 			/* We don't want the delimiters in the token */
-			token->string++;
-			real_length = string_end - token->string;
+			token->string.string++;
+			real_length = string_end - token->string.string;
 			string = string_end + 1;
 			type = CSS_TOKEN_STRING;
 		}
@@ -345,7 +345,7 @@ scan_css_token(struct dom_scanner *scanner, struct dom_scanner_token *token)
 	}
 
 	token->type = type;
-	token->length = real_length > 0 ? real_length : string - token->string;
+	token->string.length = real_length > 0 ? real_length : string - token->string.string;
 	token->precedence = get_css_precedence(type);
 	scanner->position = string;
 }
diff --git a/src/dom/scanner.h b/src/dom/scanner.h
index 9c49c756..9b247558 100644
--- a/src/dom/scanner.h
+++ b/src/dom/scanner.h
@@ -21,8 +21,7 @@ struct dom_scanner_token {
 	int precedence;
 
 	/* The start of the token string and the token length */
-	unsigned char *string;
-	int length;
+	struct dom_string string;
 };
 
 /* The naming of these two macros is a bit odd .. we compare often with
@@ -31,7 +30,7 @@ struct dom_scanner_token {
 
 /* Compare the string of @token with @string */
 #define dom_scanner_token_strlcasecmp(token, str, len) \
-	((token) && !strlcasecmp((token)->string, (token)->length, str, len))
+	((token) && !strlcasecmp((token)->string.string, (token)->string.length, str, len))
 
 /* Also compares the token string but using a "static" string */
 #define dom_scanner_token_contains(token, str) \
diff --git a/src/dom/select.c b/src/dom/select.c
index 51c79870..7bb0e640 100644
--- a/src/dom/select.c
+++ b/src/dom/select.c
@@ -70,13 +70,10 @@ get_dom_select_pseudo(struct dom_scanner_token *token)
 #undef INIT_DOM_SELECT_PSEUDO_STRING
 
 	};
-	struct dom_string string;
 	int i;
 
-	set_dom_string(&string, token->string, token->length);
-
 	for (i = 0; i < sizeof_array(pseudo_info); i++)
-		if (!dom_string_casecmp(&pseudo_info[i].string, &string))
+		if (!dom_string_casecmp(&pseudo_info[i].string, &token->string))
 			return pseudo_info[i].pseudo;
 
 	return DOM_SELECT_PSEUDO_UNKNOWN;
@@ -99,7 +96,7 @@ parse_dom_select_attribute(struct dom_select_node *sel, struct dom_scanner *scan
 	if (!token || token->type != CSS_TOKEN_IDENT)
 		return DOM_ERR_SYNTAX;
 
-	set_dom_string(&sel->node.string, token->string, token->length);
+	copy_dom_string(&sel->node.string, &token->string);
 
 	/* Get the optional '=' combo or ending ']'. */
 
@@ -143,7 +140,7 @@ parse_dom_select_attribute(struct dom_select_node *sel, struct dom_scanner *scan
 	switch (token->type) {
 	case CSS_TOKEN_IDENT:
 	case CSS_TOKEN_STRING:
-		set_dom_string(&sel->node.data.attribute.value, token->string, token->length);
+		copy_dom_string(&sel->node.data.attribute.value, &token->string);
 		break;
 
 	default:
@@ -176,7 +173,7 @@ get_scanner_token_number(struct dom_scanner_token *token)
 {
 	size_t number = 0;
 
-	while (token->length > 0 && isdigit(token->string[0])) {
+	while (token->string.length > 0 && isdigit(token->string.string[0])) {
 		size_t old_number = number;
 
 		number *= 10;
@@ -185,8 +182,8 @@ get_scanner_token_number(struct dom_scanner_token *token)
 		if (old_number > number)
 			return -1;
 
-		number += token->string[0] - '0';
-		token->string++, token->length--;
+		number += token->string.string[0] - '0';
+		token->string.string++, token->string.length--;
 	}
 
 	return number;
@@ -416,8 +413,8 @@ parse_dom_select(struct dom_select *select, struct dom_stack *stack,
 		switch (token->type) {
 		case CSS_TOKEN_IDENT:
 			sel.node.type = DOM_NODE_ELEMENT;
-			set_dom_string(&sel.node.string, token->string, token->length);
-			if (token->length == 1 && token->string[0] == '*')
+			copy_dom_string(&sel.node.string, &token->string);
+			if (token->string.length == 1 && token->string.string[0] == '*')
 				sel.match.element |= DOM_SELECT_ELEMENT_UNIVERSAL;
 			break;
 
@@ -427,7 +424,7 @@ parse_dom_select(struct dom_select *select, struct dom_stack *stack,
 			sel.node.type = DOM_NODE_ATTRIBUTE;
 			sel.match.attribute |= DOM_SELECT_ATTRIBUTE_ID;
 			/* Skip the leading '#'. */
-			token->string++, token->length--;
+			token->string.string++, token->string.length--;
 			break;
 
 		case '[':
@@ -445,7 +442,7 @@ parse_dom_select(struct dom_select *select, struct dom_stack *stack,
 			sel.node.type = DOM_NODE_ATTRIBUTE;
 			sel.match.attribute |= DOM_SELECT_ATTRIBUTE_SPACE_LIST;
 			set_dom_string(&sel.node.string, "class", -1); 
-			set_dom_string(&sel.node.data.attribute.value, token->string, token->length); 
+			copy_dom_string(&sel.node.data.attribute.value, &token->string); 
 			break;
 
 		case ':':
diff --git a/src/dom/sgml/parser.c b/src/dom/sgml/parser.c
index 98dbed1d..2fb03210 100644
--- a/src/dom/sgml/parser.c
+++ b/src/dom/sgml/parser.c
@@ -56,7 +56,7 @@ add_sgml_element(struct dom_stack *stack, struct dom_scanner_token *token)
 	struct dom_node *node;
 	struct sgml_node_info *node_info;
 
-	node = add_dom_element(parent, token->string, token->length);
+	node = add_dom_element(parent, token->string.string, token->string.length);
 	if (!node) return NULL;
 
 	node_info = get_sgml_node_info(parser->info->elements, node);
@@ -81,12 +81,12 @@ add_sgml_attribute(struct dom_stack *stack,
 {
 	struct sgml_parser *parser = get_sgml_parser(stack);
 	struct dom_node *parent = get_dom_stack_top(stack)->node;
-	unsigned char *value = valtoken ? valtoken->string : NULL;
-	size_t valuelen = valtoken ? valtoken->length : 0;
+	unsigned char *value = valtoken ? valtoken->string.string : NULL;
+	size_t valuelen = valtoken ? valtoken->string.length : 0;
 	struct sgml_node_info *info;
 	struct dom_node *node;
 
-	node = add_dom_attribute(parent, token->string, token->length,
+	node = add_dom_attribute(parent, token->string.string, token->string.length,
 				 value, valuelen);
 
 	info = get_sgml_node_info(parser->info->attributes, node);
@@ -110,15 +110,15 @@ add_sgml_proc_instruction(struct dom_stack *stack, struct dom_scanner_token *tok
 	struct dom_node *parent = get_dom_stack_top(stack)->node;
 	struct dom_node *node;
 	/* Split the token in two if we can find a first space separator. */
-	unsigned char *separator = memchr(token->string, ' ', token->length);
+	unsigned char *separator = memchr(token->string.string, ' ', token->string.length);
 
 	/* Anything before the separator becomes the target name ... */
-	unsigned char *name = token->string;
-	size_t namelen = separator ? separator - token->string : token->length;
+	unsigned char *name = token->string.string;
+	size_t namelen = separator ? separator - token->string.string : token->string.length;
 
 	/* ... and everything after the instruction value. */
 	unsigned char *value = separator ? separator + 1 : NULL;
-	size_t valuelen = value ? token->length - namelen - 1 : 0;
+	size_t valuelen = value ? token->string.length - namelen - 1 : 0;
 
 	node = add_dom_proc_instruction(parent, name, namelen, value, valuelen);
 	if (!node) return NULL;
@@ -146,7 +146,7 @@ static inline void
 add_sgml_node(struct dom_stack *stack, enum dom_node_type type, struct dom_scanner_token *token)
 {
 	struct dom_node *parent = get_dom_stack_top(stack)->node;
-	struct dom_node *node = add_dom_node(parent, type, token->string, token->length);
+	struct dom_node *node = add_dom_node(parent, type, token->string.string, token->string.length);
 
 	if (!node) return;
 
@@ -251,13 +251,13 @@ parse_sgml_plain(struct dom_stack *stack, struct dom_scanner *scanner)
 			break;
 
 		case SGML_TOKEN_ELEMENT_END:
-			if (!token->length) {
+			if (!token->string.length) {
 				pop_dom_node(stack);
 			} else {
 				struct dom_string string;
 				struct dom_stack_state *state;
 
-				set_dom_string(&string, token->string, token->length);
+				set_dom_string(&string, token->string.string, token->string.length);
 				state = search_dom_stack(stack, DOM_NODE_ELEMENT,
 							 &string);
 				if (state) {
diff --git a/src/dom/sgml/scanner.c b/src/dom/sgml/scanner.c
index ddd67559..7987a513 100644
--- a/src/dom/sgml/scanner.c
+++ b/src/dom/sgml/scanner.c
@@ -101,14 +101,14 @@ scan_sgml_text_token(struct dom_scanner *scanner, struct dom_scanner_token *toke
 	/* In scan_sgml_tokens() we check that first_char != '<' */
 	assert(first_char != '<' && scanner->state == SGML_STATE_TEXT);
 
-	token->string = string++;
+	token->string.string = string++;
 
 	if (first_char == '&') {
 		if (is_sgml_entity(*string)) {
 			scan_sgml(scanner, string, SGML_CHAR_ENTITY);
 			type = SGML_TOKEN_ENTITY;
-			token->string++;
-			real_length = string - token->string;
+			token->string.string++;
+			real_length = string - token->string.string;
 		}
 
 		foreach_sgml_cdata (scanner, string) {
@@ -133,7 +133,7 @@ scan_sgml_text_token(struct dom_scanner *scanner, struct dom_scanner_token *toke
 	}
 
 	token->type = type;
-	token->length = real_length >= 0 ? real_length : string - token->string;
+	token->string.length = real_length >= 0 ? real_length : string - token->string.string;
 	token->precedence = get_sgml_precedence(type);
 	scanner->position = string;
 }
@@ -223,7 +223,7 @@ scan_sgml_element_token(struct dom_scanner *scanner, struct dom_scanner_token *t
 	enum sgml_token_type type = SGML_TOKEN_GARBAGE;
 	int real_length = -1;
 
-	token->string = string++;
+	token->string.string = string++;
 
 	if (first_char == '<') {
 		scan_sgml(scanner, string, SGML_CHAR_WHITESPACE);
@@ -237,10 +237,10 @@ scan_sgml_element_token(struct dom_scanner *scanner, struct dom_scanner_token *t
 			scanner->state = SGML_STATE_TEXT;
 
 		} else if (is_sgml_ident(*string)) {
-			token->string = string;
+			token->string.string = string;
 			scan_sgml(scanner, string, SGML_CHAR_IDENT);
 
-			real_length = string - token->string;
+			real_length = string - token->string.string;
 
 			scan_sgml(scanner, string, SGML_CHAR_WHITESPACE);
 			if (*string == '>') {
@@ -257,13 +257,13 @@ scan_sgml_element_token(struct dom_scanner *scanner, struct dom_scanner_token *t
 
 			string++;
 			scan_sgml(scanner, string, SGML_CHAR_WHITESPACE);
-			token->string = ident = string;
+			token->string.string = ident = string;
 
 			if (string + 1 < scanner->end
 			    && string[0] == '-' && string[1] == '-') {
 				string += 2;
 				type = SGML_TOKEN_NOTATION_COMMENT;
-				token->string = string;
+				token->string.string = string;
 				real_length = skip_comment(scanner, &string);
 				assert(real_length >= 0);
 
@@ -272,7 +272,7 @@ scan_sgml_element_token(struct dom_scanner *scanner, struct dom_scanner_token *t
 
 				string += 7;
 				type = SGML_TOKEN_CDATA_SECTION;
-				token->string = string;
+				token->string.string = string;
 				real_length = skip_cdata_section(scanner, &string);
 				assert(real_length >= 0);
 
@@ -288,7 +288,7 @@ scan_sgml_element_token(struct dom_scanner *scanner, struct dom_scanner_token *t
 
 			string++;
 			scan_sgml(scanner, string, SGML_CHAR_WHITESPACE);
-			token->string = pos = string;
+			token->string.string = pos = string;
 			scan_sgml(scanner, string, SGML_CHAR_IDENT);
 
 			type = map_dom_scanner_string(scanner, pos, string, base);
@@ -299,7 +299,7 @@ scan_sgml_element_token(struct dom_scanner *scanner, struct dom_scanner_token *t
 
 				/* Set length until '?' char and move position
 				 * beyond '>'. */
-				real_length = pos - token->string - 2;
+				real_length = pos - token->string.string - 2;
 				break;
 			}
 
@@ -320,9 +320,9 @@ scan_sgml_element_token(struct dom_scanner *scanner, struct dom_scanner_token *t
 			scan_sgml(scanner, string, SGML_CHAR_WHITESPACE);
 
 			if (is_sgml_ident(*string)) {
-				token->string = string;
+				token->string.string = string;
 				scan_sgml(scanner, string, SGML_CHAR_IDENT);
-				real_length = string - token->string;
+				real_length = string - token->string.string;
 
 				type = SGML_TOKEN_ELEMENT_END;
 				skip_sgml(scanner, &string, '>', 1);
@@ -371,12 +371,12 @@ scan_sgml_element_token(struct dom_scanner *scanner, struct dom_scanner_token *t
 
 		if (string_end) {
 			/* We don't want the delimiters in the token */
-			token->string++;
-			real_length = string_end - token->string;
+			token->string.string++;
+			real_length = string_end - token->string.string;
 			string = string_end + 1;
 			type = SGML_TOKEN_STRING;
 		} else if (is_sgml_attribute(*string)) {
-			token->string++;
+			token->string.string++;
 			scan_sgml_attribute(scanner, string);
 			type = SGML_TOKEN_ATTRIBUTE;
 		}
@@ -394,7 +394,7 @@ scan_sgml_element_token(struct dom_scanner *scanner, struct dom_scanner_token *t
 	}
 
 	token->type = type;
-	token->length = real_length >= 0 ? real_length : string - token->string;
+	token->string.length = real_length >= 0 ? real_length : string - token->string.string;
 	token->precedence = get_sgml_precedence(type);
 	scanner->position = string;
 }
diff --git a/src/dom/string.h b/src/dom/string.h
index ba543a76..50030e4e 100644
--- a/src/dom/string.h
+++ b/src/dom/string.h
@@ -35,6 +35,9 @@ dom_string_ncasecmp(struct dom_string *string1, struct dom_string *string2, size
 	return strncasecmp(string1->string, string2->string, length);
 }
 
+#define copy_dom_string(string1, string2) \
+	set_dom_string(string1, (string2)->string, (string2)->length)
+
 static inline struct dom_string *
 init_dom_string(struct dom_string *string, unsigned char *str, size_t len)
 {