diff --git a/src/pkg/html/node.go b/src/pkg/html/node.go
index 595afd5690..4ecfd6ca23 100644
--- a/src/pkg/html/node.go
+++ b/src/pkg/html/node.go
@@ -13,6 +13,7 @@ const (
 	DocumentNode
 	ElementNode
 	CommentNode
+	DoctypeNode
 	scopeMarkerNode
 )
 
diff --git a/src/pkg/html/parse.go b/src/pkg/html/parse.go
index 980c470691..519ebe587b 100644
--- a/src/pkg/html/parse.go
+++ b/src/pkg/html/parse.go
@@ -81,8 +81,8 @@ func (p *parser) popUntil(stopTags []string, matchTags ...string) bool {
 	return false
 }
 
-// addChild adds a child node n to the top element, and pushes n if it is an
-// element node (text nodes are not part of the stack of open elements).
+// addChild adds a child node n to the top element, and pushes n onto the stack
+// of open elements if it is an element node.
 func (p *parser) addChild(n *Node) {
 	p.top().Add(n)
 	if n.Type == ElementNode {
@@ -90,10 +90,15 @@ func (p *parser) addChild(n *Node) {
 	}
 }
 
-// addText calls addChild with a text node.
+// addText adds text to the preceding node if it is a text node, or else it
+// calls addChild with a new text node.
 func (p *parser) addText(text string) {
-	// TODO: merge s with previous text, if the preceding node is a text node.
 	// TODO: distinguish whitespace text from others.
+	t := p.top()
+	if i := len(t.Child); i > 0 && t.Child[i-1].Type == TextNode {
+		t.Child[i-1].Data += text
+		return
+	}
 	p.addChild(&Node{
 		Type: TextNode,
 		Data: text,
@@ -201,7 +206,15 @@ func useTheRulesFor(p *parser, actual, delegate insertionMode) (insertionMode, b
 
 // Section 11.2.5.4.1.
 func initialIM(p *parser) (insertionMode, bool) {
-	// TODO: check p.tok for DOCTYPE.
+	if p.tok.Type == DoctypeToken {
+		p.addChild(&Node{
+			Type: DoctypeNode,
+			Data: p.tok.Data,
+		})
+		return beforeHTMLIM, true
+	}
+	// TODO: set "quirks mode"? It's defined in the DOM spec instead of HTML5 proper,
+	// and so switching on "quirks mode" might belong in a different package.
 	return beforeHTMLIM, false
 }
 
diff --git a/src/pkg/html/parse_test.go b/src/pkg/html/parse_test.go
index f22fa277bc..7d918d2508 100644
--- a/src/pkg/html/parse_test.go
+++ b/src/pkg/html/parse_test.go
@@ -85,6 +85,8 @@ func dumpLevel(w io.Writer, n *Node, level int) os.Error {
 		fmt.Fprintf(w, "%q", EscapeString(n.Data))
 	case CommentNode:
 		return os.NewError("COMMENT")
+	case DoctypeNode:
+		fmt.Fprintf(w, "<!DOCTYPE %s>", EscapeString(n.Data))
 	case scopeMarkerNode:
 		return os.NewError("unexpected scopeMarkerNode")
 	default:
@@ -121,7 +123,7 @@ func TestParser(t *testing.T) {
 		rc := make(chan io.Reader)
 		go readDat(filename, rc)
 		// TODO(nigeltao): Process all test cases, not just a subset.
-		for i := 0; i < 23; i++ {
+		for i := 0; i < 25; i++ {
 			// Parse the #data section.
 			b, err := ioutil.ReadAll(<-rc)
 			if err != nil {
diff --git a/src/pkg/html/token.go b/src/pkg/html/token.go
index 5c6ed16662..d280ff2256 100644
--- a/src/pkg/html/token.go
+++ b/src/pkg/html/token.go
@@ -27,6 +27,8 @@ const (
 	SelfClosingTagToken
 	// A CommentToken looks like <!--x-->.
 	CommentToken
+	// A DoctypeToken looks like <!DOCTYPE x>
+	DoctypeToken
 )
 
 // String returns a string representation of the TokenType.
@@ -44,6 +46,8 @@ func (t TokenType) String() string {
 		return "SelfClosingTag"
 	case CommentToken:
 		return "Comment"
+	case DoctypeToken:
+		return "Doctype"
 	}
 	return "Invalid(" + strconv.Itoa(int(t)) + ")"
 }
@@ -56,9 +60,9 @@ type Attribute struct {
 }
 
 // A Token consists of a TokenType and some Data (tag name for start and end
-// tags, content for text and comments). A tag Token may also contain a slice
-// of Attributes. Data is unescaped for all Tokens (it looks like "a<b" rather
-// than "a&lt;b").
+// tags, content for text, comments and doctypes). A tag Token may also contain
+// a slice of Attributes. Data is unescaped for all Tokens (it looks like "a<b"
+// rather than "a&lt;b").
 type Token struct {
 	Type TokenType
 	Data string
@@ -97,6 +101,8 @@ func (t Token) String() string {
 		return "<" + t.tagString() + "/>"
 	case CommentToken:
 		return "<!--" + EscapeString(t.Data) + "-->"
+	case DoctypeToken:
+		return "<!DOCTYPE " + EscapeString(t.Data) + ">"
 	}
 	return "Invalid(" + strconv.Itoa(int(t.Type)) + ")"
 }
@@ -109,9 +115,15 @@ type Tokenizer struct {
 
 	// r is the source of the HTML text.
 	r io.Reader
-	// tt is the TokenType of the most recently read token. If tt == Error
-	// then err is the error associated with trying to read that token.
-	tt  TokenType
+	// tt is the TokenType of the most recently read token.
+	tt TokenType
+	// err is the first error encountered during tokenization. It is possible
+	// for tt != Error && err != nil to hold: this means that Next returned a
+	// valid token but the subsequent Next call will return an error token.
+	// For example, if the HTML text input was just "plain", then the first
+	// Next call would set z.err to os.EOF but return a TextToken, and all
+	// subsequent Next calls would return an ErrorToken.
+	// err is never reset. Once it becomes non-nil, it stays non-nil.
 	err os.Error
 	// buf[p0:p1] holds the raw data of the most recent token.
 	// buf[p1:] is buffered input that will yield future tokens.
@@ -137,7 +149,9 @@ func (z *Tokenizer) Raw() []byte {
 // readByte returns the next byte from the input stream, doing a buffered read
 // from z.r into z.buf if necessary. z.buf[z.p0:z.p1] remains a contiguous byte
 // slice that holds all the bytes read so far for the current token.
-func (z *Tokenizer) readByte() (byte, os.Error) {
+// It sets z.err if the underlying reader returns an error.
+// Pre-condition: z.err == nil.
+func (z *Tokenizer) readByte() byte {
 	if z.p1 >= len(z.buf) {
 		// Our buffer is exhausted and we have to read from z.r.
 		// We copy z.buf[z.p0:z.p1] to the beginning of z.buf. If the length
@@ -149,82 +163,120 @@ func (z *Tokenizer) readByte() (byte, os.Error) {
 		if 2*d > c {
 			buf1 = make([]byte, d, 2*c)
 		} else {
-			buf1 = z.buf[0:d]
+			buf1 = z.buf[:d]
 		}
 		copy(buf1, z.buf[z.p0:z.p1])
-		z.p0, z.p1, z.buf = 0, d, buf1[0:d]
+		z.p0, z.p1, z.buf = 0, d, buf1[:d]
 		// Now that we have copied the live bytes to the start of the buffer,
 		// we read from z.r into the remainder.
 		n, err := z.r.Read(buf1[d:cap(buf1)])
 		if err != nil {
-			return 0, err
+			z.err = err
+			return 0
 		}
-		z.buf = buf1[0 : d+n]
+		z.buf = buf1[:d+n]
 	}
 	x := z.buf[z.p1]
 	z.p1++
-	return x, nil
+	return x
 }
 
-// readTo keeps reading bytes until x is found.
-func (z *Tokenizer) readTo(x uint8) os.Error {
+// readTo keeps reading bytes until x is found or a read error occurs. If an
+// error does occur, z.err is set to that error.
+// Pre-condition: z.err == nil.
+func (z *Tokenizer) readTo(x uint8) {
 	for {
-		c, err := z.readByte()
-		if err != nil {
-			return err
+		c := z.readByte()
+		if z.err != nil {
+			return
 		}
 		switch c {
 		case x:
-			return nil
+			return
 		case '\\':
-			_, err = z.readByte()
-			if err != nil {
-				return err
+			z.readByte()
+			if z.err != nil {
+				return
 			}
 		}
 	}
-	panic("unreachable")
 }
 
-// nextMarkupDeclaration returns the next TokenType starting with "<!".
-func (z *Tokenizer) nextMarkupDeclaration() (TokenType, os.Error) {
-	// TODO: check for <!DOCTYPE ... >, don't just assume that it's a comment.
-	for i := 0; i < 2; i++ {
-		c, err := z.readByte()
-		if err != nil {
-			return TextToken, err
-		}
-		if c != '-' {
-			return z.nextText(), nil
-		}
-	}
+// nextComment reads the next token starting with "<!--".
+// The opening "<!--" has already been consumed.
+// Pre-condition: z.tt == TextToken && z.err == nil && z.p0 + 4 <= z.p1.
+func (z *Tokenizer) nextComment() {
 	// <!--> is a valid comment.
 	for dashCount := 2; ; {
-		c, err := z.readByte()
-		if err != nil {
-			return TextToken, err
+		c := z.readByte()
+		if z.err != nil {
+			return
 		}
 		switch c {
 		case '-':
 			dashCount++
 		case '>':
 			if dashCount >= 2 {
-				return CommentToken, nil
+				z.tt = CommentToken
+				return
 			}
-			fallthrough
+			dashCount = 0
 		default:
 			dashCount = 0
 		}
 	}
-	panic("unreachable")
 }
 
-// nextTag returns the next TokenType starting from the tag open state.
-func (z *Tokenizer) nextTag() (tt TokenType, err os.Error) {
-	c, err := z.readByte()
-	if err != nil {
-		return ErrorToken, err
+// nextMarkupDeclaration reads the next token starting with "<!".
+// It might be a "<!--comment-->", a "<!DOCTYPE foo>", or "<!malformed text".
+// The opening "<!" has already been consumed.
+// Pre-condition: z.tt == TextToken && z.err == nil && z.p0 + 2 <= z.p1.
+func (z *Tokenizer) nextMarkupDeclaration() {
+	var c [2]byte
+	for i := 0; i < 2; i++ {
+		c[i] = z.readByte()
+		if z.err != nil {
+			return
+		}
 	}
+	if c[0] == '-' && c[1] == '-' {
+		z.nextComment()
+		return
+	}
+	z.p1 -= 2
+	const s = "DOCTYPE "
+	for i := 0; ; i++ {
+		c := z.readByte()
+		if z.err != nil {
+			return
+		}
+		// Capitalize c.
+		if 'a' <= c && c <= 'z' {
+			c = 'A' + (c - 'a')
+		}
+		if i < len(s) && c != s[i] {
+			z.nextText()
+			return
+		}
+		if c == '>' {
+			if i >= len(s) {
+				z.tt = DoctypeToken
+			}
+			return
+		}
+	}
+}
+
+// nextTag reads the next token starting with "<". It might be a "<startTag>",
+// an "</endTag>", a "<!markup declaration>", or "<malformed text".
+// The opening "<" has already been consumed.
+// Pre-condition: z.tt == TextToken && z.err == nil && z.p0 + 1 <= z.p1.
+func (z *Tokenizer) nextTag() {
+	c := z.readByte()
+	if z.err != nil {
+		return
+	}
+	var tt TokenType
 	switch {
 	case c == '/':
 		tt = EndTagToken
@@ -232,56 +284,49 @@ func (z *Tokenizer) nextTag() (tt TokenType, err os.Error) {
 	case 'a' <= c && c <= 'z' || 'A' <= c && c <= 'Z':
 		tt = StartTagToken
 	case c == '!':
-		return z.nextMarkupDeclaration()
+		z.nextMarkupDeclaration()
+		return
 	case c == '?':
-		return ErrorToken, os.NewError("html: TODO(nigeltao): implement XML processing instructions")
+		z.tt, z.err = ErrorToken, os.NewError("html: TODO: implement XML processing instructions")
+		return
 	default:
-		return ErrorToken, os.NewError("html: TODO(nigeltao): handle malformed tags")
+		z.tt, z.err = ErrorToken, os.NewError("html: TODO: handle malformed tags")
+		return
 	}
 	for {
-		c, err := z.readByte()
-		if err != nil {
-			return TextToken, err
+		c := z.readByte()
+		if z.err != nil {
+			return
 		}
 		switch c {
-		case '"':
-			err = z.readTo('"')
-			if err != nil {
-				return TextToken, err
-			}
-		case '\'':
-			err = z.readTo('\'')
-			if err != nil {
-				return TextToken, err
+		case '"', '\'':
+			z.readTo(c)
+			if z.err != nil {
+				return
 			}
 		case '>':
+			z.tt = tt
 			if z.buf[z.p1-2] == '/' && tt == StartTagToken {
-				return SelfClosingTagToken, nil
+				z.tt = SelfClosingTagToken
 			}
-			return tt, nil
+			return
 		}
 	}
-	panic("unreachable")
 }
 
 // nextText reads all text up until an '<'.
-func (z *Tokenizer) nextText() TokenType {
+// Pre-condition: z.tt == TextToken && z.err == nil && z.p0 + 1 <= z.p1.
+func (z *Tokenizer) nextText() {
 	for {
-		c, err := z.readByte()
-		if err != nil {
-			z.tt, z.err = ErrorToken, err
-			if err == os.EOF {
-				z.tt = TextToken
-			}
-			return z.tt
+		c := z.readByte()
+		if z.err != nil {
+			return
 		}
 		if c == '<' {
 			z.p1--
-			z.tt = TextToken
-			return z.tt
+			return
 		}
 	}
-	panic("unreachable")
 }
 
 // Next scans the next token and returns its type.
@@ -292,19 +337,22 @@ func (z *Tokenizer) Next() TokenType {
 			return z.tt
 		}
 		z.p0 = z.p1
-		c, err := z.readByte()
-		if err != nil {
-			z.tt, z.err = ErrorToken, err
+		c := z.readByte()
+		if z.err != nil {
+			z.tt = ErrorToken
 			return z.tt
 		}
-		if c == '<' {
-			z.tt, z.err = z.nextTag()
+		// We assume that the next token is text unless proven otherwise.
+		z.tt = TextToken
+		if c != '<' {
+			z.nextText()
+		} else {
+			z.nextTag()
 			if z.tt == CommentToken && !z.ReturnComments {
 				continue
 			}
-			return z.tt
 		}
-		return z.nextText()
+		return z.tt
 	}
 	panic("unreachable")
 }
@@ -382,25 +430,29 @@ loop:
 	return z.buf[i0:i], z.trim(i)
 }
 
-// Text returns the unescaped text of a TextToken or a CommentToken.
-// The contents of the returned slice may change on the next call to Next.
+// Text returns the unescaped text of a text, comment or doctype token. The
+// contents of the returned slice may change on the next call to Next.
 func (z *Tokenizer) Text() []byte {
+	var i0, i1 int
 	switch z.tt {
 	case TextToken:
-		s := unescape(z.Raw())
-		z.p0 = z.p1
-		return s
+		i0 = z.p0
+		i1 = z.p1
 	case CommentToken:
-		// We trim the "<!--" from the left and the "-->" from the right.
+		// Trim the "<!--" from the left and the "-->" from the right.
 		// "<!-->" is a valid comment, so the adjusted endpoints might overlap.
-		i0 := z.p0 + 4
-		i1 := z.p1 - 3
-		z.p0 = z.p1
-		var s []byte
-		if i0 < i1 {
-			s = unescape(z.buf[i0:i1])
-		}
-		return s
+		i0 = z.p0 + 4
+		i1 = z.p1 - 3
+	case DoctypeToken:
+		// Trim the "<!DOCTYPE " from the left and the ">" from the right.
+		i0 = z.p0 + 10
+		i1 = z.p1 - 1
+	default:
+		return nil
+	}
+	z.p0 = z.p1
+	if i0 < i1 {
+		return unescape(z.buf[i0:i1])
 	}
 	return nil
 }
@@ -483,7 +535,7 @@ loop:
 func (z *Tokenizer) Token() Token {
 	t := Token{Type: z.tt}
 	switch z.tt {
-	case TextToken, CommentToken:
+	case TextToken, CommentToken, DoctypeToken:
 		t.Data = string(z.Text())
 	case StartTagToken, EndTagToken, SelfClosingTagToken:
 		var attr []Attribute