work on stream

stream/stream.go

@@ -2,6 +2,7 @@ package stream
 
 import (
 	"encoding/binary"
+	"fmt"
 	"sync"
 	"sync/atomic"
 	"time"
@@ -26,20 +27,25 @@ type Stream struct {
 	decodeKey int
 
 	// Sequence management
-	seqOut     uint16
-	seqIn      uint16
-	seqLastAck uint16
+	seqOut      uint16
+	seqIn       uint16
+	seqLastAck  uint16
+	seqExpected uint16
 
 	// Acknowledgment tracking
 	pendingAcks  map[uint16]*pendingPacket
 	ackTimer     *time.Timer
 	lastAckSent  time.Time
 	ackThreshold time.Duration
+	ackQueue     []uint16 // Sequences needing ACK
 
 	// Fragment assembly
 	fragments  map[uint16]*fragmentBuffer
 	nextFragID uint16
 
+	// Out of order handling
+	outOfOrder map[uint16][]byte
+
 	// Packet queues
 	reliableQueue   []*packets.ProtoPacket
 	unreliableQueue []*packets.ProtoPacket
@@ -56,11 +62,20 @@ type Stream struct {
 	keepAliveTimer *time.Timer
 	timeoutTimer   *time.Timer
 
+	// Retransmission settings
+	rtt        time.Duration
+	rttVar     time.Duration
+	rto        time.Duration
+	minRTO     time.Duration
+	maxRTO     time.Duration
+	maxRetries int
+
 	// Stats
-	packetsOut uint64
-	packetsIn  uint64
-	bytesOut   uint64
-	bytesIn    uint64
+	packetsOut  uint64
+	packetsIn   uint64
+	bytesOut    uint64
+	bytesIn     uint64
+	retransmits uint64
 
 	// Callbacks
 	onPacket     func(*packets.AppPacket)
@@ -120,8 +135,17 @@ func NewStream(conn gnet.Conn, cfg *Config) *Stream {
 		ackThreshold:  cfg.AckThreshold,
 		pendingAcks:   make(map[uint16]*pendingPacket),
 		fragments:     make(map[uint16]*fragmentBuffer),
+		outOfOrder:    make(map[uint16][]byte),
+		ackQueue:      make([]uint16, 0),
 		seqOut:        0,
 		seqIn:         0,
+		seqExpected:   0,
+		rtt:           time.Second,
+		rttVar:        500 * time.Millisecond,
+		rto:           time.Second,
+		minRTO:        200 * time.Millisecond,
+		maxRTO:        10 * time.Second,
+		maxRetries:    10,
 	}
 
 	if s.maxLen == 0 {
@@ -144,6 +168,11 @@ func NewStream(conn gnet.Conn, cfg *Config) *Stream {
 		s.keepAliveTimer = time.AfterFunc(cfg.KeepaliveTime, s.sendKeepalive)
 	}
 
+	// Start timeout timer
+	if cfg.TimeoutDuration > 0 {
+		s.timeoutTimer = time.AfterFunc(cfg.TimeoutDuration, s.handleTimeout)
+	}
+
 	return s
 }
 
@@ -195,6 +224,8 @@ func (s *Stream) Process(data []byte) error {
 		return nil
 	case opcodes.OP_SessionDisconnect:
 		return s.handleDisconnect()
+	case opcodes.OP_OutOfOrderAck:
+		return s.handleOutOfOrderAck(data[2:])
 	}
 
 	atomic.AddUint64(&s.packetsIn, 1)
@@ -203,6 +234,62 @@ func (s *Stream) Process(data []byte) error {
 	return nil
 }
 
+// handleSessionRequest processes session establishment request
+func (s *Stream) handleSessionRequest(data []byte) error {
+	if len(data) < 10 {
+		return fmt.Errorf("session request too small")
+	}
+
+	// Parse session request
+	// Format: version(4) + sessionID(4) + maxLen(2)
+	version := binary.BigEndian.Uint32(data[:4])
+	sessionID := binary.BigEndian.Uint32(data[4:8])
+	maxLen := binary.BigEndian.Uint16(data[8:10])
+
+	// Update session info
+	s.mu.Lock()
+	s.sessionID = sessionID
+	if maxLen < s.maxLen {
+		s.maxLen = maxLen // Use smaller of the two
+	}
+	s.mu.Unlock()
+
+	// Send session response
+	response := make([]byte, 14)
+	binary.BigEndian.PutUint32(response[0:4], sessionID)
+	binary.BigEndian.PutUint32(response[4:8], s.crcKey)
+	response[8] = 2 // Encoding type
+	binary.BigEndian.PutUint16(response[9:11], s.maxLen)
+	binary.BigEndian.PutUint32(response[11:15], version)
+
+	s.state.Store(StateEstablished)
+	return s.sendRaw(opcodes.OP_SessionResponse, response)
+}
+
+// handleSessionResponse processes session establishment response
+func (s *Stream) handleSessionResponse(data []byte) error {
+	if len(data) < 14 {
+		return fmt.Errorf("session response too small")
+	}
+
+	// Parse response
+	sessionID := binary.BigEndian.Uint32(data[0:4])
+	crcKey := binary.BigEndian.Uint32(data[4:8])
+	// encodingType := data[8]
+	maxLen := binary.BigEndian.Uint16(data[9:11])
+
+	s.mu.Lock()
+	s.sessionID = sessionID
+	s.crcKey = crcKey
+	if maxLen < s.maxLen {
+		s.maxLen = maxLen
+	}
+	s.mu.Unlock()
+
+	s.state.Store(StateEstablished)
+	return nil
+}
+
 // handlePacket processes reliable packets
 func (s *Stream) handlePacket(data []byte) error {
 	if len(data) < 2 {
@@ -215,47 +302,70 @@ func (s *Stream) handlePacket(data []byte) error {
 	s.mu.Lock()
 	defer s.mu.Unlock()
 
-	// Check sequence
-	if seq != s.seqIn {
-		// Out of order - send immediate ACK
-		s.sendAckImmediate(seq)
-		return nil
+	// Check if this is the expected sequence
+	if seq == s.seqExpected {
+		// Process in-order packet
+		s.seqExpected++
+		s.ackQueue = append(s.ackQueue, seq)
+
+		// Process packet data
+		if err := s.processPacketData(data); err != nil {
+			return err
+		}
+
+		// Check for buffered out-of-order packets
+		for {
+			if buffered, exists := s.outOfOrder[s.seqExpected]; exists {
+				delete(s.outOfOrder, s.seqExpected)
+				s.ackQueue = append(s.ackQueue, s.seqExpected)
+				s.seqExpected++
+				if err := s.processPacketData(buffered); err != nil {
+					return err
+				}
+			} else {
+				break
+			}
+		}
+	} else if seq > s.seqExpected {
+		// Out of order - buffer it
+		s.outOfOrder[seq] = append([]byte(nil), data...)
+		// Send out-of-order ACK
+		go s.sendOutOfOrderAck(seq)
+	} else {
+		// Duplicate packet - just ACK it
+		go s.sendAckImmediate(seq)
 	}
 
-	// Update sequence
-	s.seqIn++
+	// Schedule grouped ACK
+	s.scheduleAck()
 
-	// Schedule ACK
-	s.scheduleAck(seq)
-
-	// Process packet data
-	return s.processPacketData(data)
+	return nil
 }
 
 // handleFragment assembles fragmented packets
 func (s *Stream) handleFragment(data []byte) error {
-	if len(data) < 8 {
+	if len(data) < 10 {
 		return nil
 	}
 
 	seq := binary.BigEndian.Uint16(data[:2])
-	fragSeq := binary.BigEndian.Uint16(data[2:4])
-	fragTotal := binary.BigEndian.Uint16(data[4:6])
-	fragCur := binary.BigEndian.Uint16(data[6:8])
-	data = data[8:]
+	fragID := binary.BigEndian.Uint32(data[2:6])
+	fragTotal := binary.BigEndian.Uint16(data[6:8])
+	fragCur := binary.BigEndian.Uint16(data[8:10])
+	data = data[10:]
 
 	s.mu.Lock()
 	defer s.mu.Unlock()
 
 	// Get or create fragment buffer
-	frag, exists := s.fragments[fragSeq]
+	frag, exists := s.fragments[uint16(fragID)]
 	if !exists {
 		frag = &fragmentBuffer{
 			totalSize: uint32(fragTotal),
 			chunks:    make(map[uint16][]byte),
 			startTime: time.Now(),
 		}
-		s.fragments[fragSeq] = frag
+		s.fragments[uint16(fragID)] = frag
 	}
 
 	// Store chunk
@@ -264,21 +374,25 @@ func (s *Stream) handleFragment(data []byte) error {
 
 	// Check if complete
 	if frag.received == uint32(fragTotal) {
-		// Reassemble
+		// Reassemble in order
 		complete := make([]byte, 0)
 		for i := uint16(0); i < fragTotal; i++ {
 			if chunk, ok := frag.chunks[i]; ok {
 				complete = append(complete, chunk...)
+			} else {
+				// Missing chunk - wait for retransmit
+				return nil
 			}
 		}
-		delete(s.fragments, fragSeq)
+		delete(s.fragments, uint16(fragID))
 
 		// Process reassembled packet
 		return s.processPacketData(complete)
 	}
 
-	// Schedule ACK for fragment
-	s.scheduleAck(seq)
+	// ACK the fragment
+	s.ackQueue = append(s.ackQueue, seq)
+	s.scheduleAck()
 
 	return nil
 }
@@ -294,16 +408,138 @@ func (s *Stream) handleAck(data []byte) error {
 	s.mu.Lock()
 	defer s.mu.Unlock()
 
-	// Remove from pending
+	// Remove from pending and update RTT
 	if pending, exists := s.pendingAcks[ackSeq]; exists {
 		delete(s.pendingAcks, ackSeq)
-		// Update RTT stats here if needed
-		_ = pending
+
+		// Update RTT estimates (TCP-like algorithm)
+		sample := time.Since(pending.sentTime)
+		if s.rtt == 0 {
+			s.rtt = sample
+			s.rttVar = sample / 2
+		} else {
+			alpha := 0.125
+			beta := 0.25
+			s.rttVar = time.Duration((1-beta)*float64(s.rttVar) + beta*float64(absTime(s.rtt-sample)))
+			s.rtt = time.Duration((1-alpha)*float64(s.rtt) + alpha*float64(sample))
+		}
+		s.rto = s.rtt + 4*s.rttVar
+		if s.rto < s.minRTO {
+			s.rto = s.minRTO
+		}
+		if s.rto > s.maxRTO {
+			s.rto = s.maxRTO
+		}
 	}
 
-	// Update last ACK seen
+	// Fast retransmit: if ACK is higher than pending packets, trigger retransmit
 	if ackSeq > s.seqLastAck {
 		s.seqLastAck = ackSeq
+
+		// Check for gaps that need immediate retransmit
+		for seq, pending := range s.pendingAcks {
+			if seq < ackSeq {
+				// This packet was likely lost, retransmit immediately
+				s.resendQueue = append(s.resendQueue, pending)
+			}
+		}
+	}
+
+	return nil
+}
+
+// handleOutOfOrderAck handles out-of-order acknowledgments
+func (s *Stream) handleOutOfOrderAck(data []byte) error {
+	if len(data) < 2 {
+		return nil
+	}
+
+	seq := binary.BigEndian.Uint16(data[:2])
+
+	s.mu.Lock()
+	defer s.mu.Unlock()
+
+	// Immediately retransmit if we have this packet pending
+	if pending, exists := s.pendingAcks[seq]; exists {
+		s.resendQueue = append(s.resendQueue, pending)
+	}
+
+	return nil
+}
+
+// handleCombined processes combined protocol packets
+func (s *Stream) handleCombined(data []byte) error {
+	pos := 0
+
+	for pos < len(data) {
+		if pos+1 > len(data) {
+			break
+		}
+
+		// Read packet size
+		size := uint16(data[pos])
+		pos++
+
+		// Check for oversized marker
+		if size == 0xff {
+			if pos+2 > len(data) {
+				break
+			}
+			size = binary.BigEndian.Uint16(data[pos : pos+2])
+			pos += 2
+		}
+
+		// Extract packet
+		if pos+int(size) > len(data) {
+			break
+		}
+
+		packet := data[pos : pos+int(size)]
+		pos += int(size)
+
+		// Process each sub-packet
+		if err := s.Process(packet); err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
+// handleAppCombined processes combined application packets
+func (s *Stream) handleAppCombined(data []byte) error {
+	pos := 0
+
+	for pos < len(data) {
+		if pos+1 > len(data) {
+			break
+		}
+
+		// Read packet size
+		size := uint16(data[pos])
+		pos++
+
+		// Check for oversized marker
+		if size == 0xff {
+			if pos+2 > len(data) {
+				break
+			}
+			size = binary.BigEndian.Uint16(data[pos : pos+2])
+			pos += 2
+		}
+
+		// Extract packet
+		if pos+int(size) > len(data) {
+			break
+		}
+
+		packet := data[pos : pos+int(size)]
+		pos += int(size)
+
+		// Process as application packet
+		if err := s.processPacketData(packet); err != nil {
+			return err
+		}
 	}
 
 	return nil
@@ -311,29 +547,79 @@ func (s *Stream) handleAck(data []byte) error {
 
 // SendPacket sends an application packet
 func (s *Stream) SendPacket(app *packets.AppPacket) error {
+	// Check state
+	if s.state.Load() != StateEstablished {
+		return fmt.Errorf("stream not established")
+	}
+
 	// Convert to protocol packet
 	proto := s.appToProto(app)
 
 	// Check if needs fragmentation
-	if proto.Size() > uint32(s.maxLen) {
+	if proto.Size() > uint32(s.maxLen-10) { // Reserve space for headers
 		return s.sendFragmented(proto)
 	}
 
-	// Add to appropriate queue
+	// Add to appropriate queue based on reliability
+	s.mu.Lock()
 	if app.Priority > packets.PriorityNormal {
-		s.mu.Lock()
 		s.reliableQueue = append(s.reliableQueue, proto)
-		s.mu.Unlock()
 	} else {
-		s.mu.Lock()
 		s.unreliableQueue = append(s.unreliableQueue, proto)
-		s.mu.Unlock()
 	}
+	s.mu.Unlock()
 
 	// Process queues
 	return s.processQueues()
 }
 
+// sendFragmented fragments and sends large packets
+func (s *Stream) sendFragmented(proto *packets.ProtoPacket) error {
+	data := make([]byte, proto.Size())
+	proto.Serialize(data, 0)
+
+	// Calculate fragment size (leave room for headers)
+	fragSize := int(s.maxLen) - 12 // Fragment header is 10 bytes + 2 for protocol
+	numFrags := (len(data) + fragSize - 1) / fragSize
+
+	if numFrags > 0xFFFF {
+		return fmt.Errorf("packet too large to fragment")
+	}
+
+	s.mu.Lock()
+	fragID := s.nextFragID
+	s.nextFragID++
+	s.mu.Unlock()
+
+	// Send each fragment
+	for i := 0; i < numFrags; i++ {
+		start := i * fragSize
+		end := start + fragSize
+		if end > len(data) {
+			end = len(data)
+		}
+
+		// Build fragment header
+		fragHeader := make([]byte, 10)
+		// Sequence added by sendReliable
+		binary.BigEndian.PutUint32(fragHeader[0:4], uint32(fragID))
+		binary.BigEndian.PutUint16(fragHeader[4:6], uint16(numFrags))
+		binary.BigEndian.PutUint16(fragHeader[6:8], uint16(i))
+		binary.BigEndian.PutUint16(fragHeader[8:10], uint16(end-start))
+
+		// Combine header and data
+		fragment := append(fragHeader, data[start:end]...)
+
+		// Send as reliable packet
+		fragProto := packets.NewProtoPacket(opcodes.OP_Fragment, fragment, s.opcodeManager)
+		if err := s.sendReliable(fragProto); err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
 // sendReliable sends a reliable packet with sequence number
 func (s *Stream) sendReliable(proto *packets.ProtoPacket) error {
 	s.mu.Lock()
@@ -348,11 +634,11 @@ func (s *Stream) sendReliable(proto *packets.ProtoPacket) error {
 
 	// Track for retransmission
 	pending := &pendingPacket{
-		packet:    proto,
+		packet:    proto.Copy(),
 		seq:       seq,
 		sentTime:  time.Now(),
 		attempts:  1,
-		nextRetry: time.Now().Add(time.Second),
+		nextRetry: time.Now().Add(s.rto),
 	}
 
 	s.mu.Lock()
@@ -363,6 +649,183 @@ func (s *Stream) sendReliable(proto *packets.ProtoPacket) error {
 	return s.sendRaw(opcodes.OP_Packet, data)
 }
 
+// processQueues processes all packet queues
+func (s *Stream) processQueues() error {
+	s.mu.Lock()
+
+	// Process resends first (highest priority)
+	for len(s.resendQueue) > 0 {
+		pending := s.resendQueue[0]
+		s.resendQueue = s.resendQueue[1:]
+
+		if time.Now().After(pending.nextRetry) {
+			s.mu.Unlock()
+
+			// Rebuild packet
+			data := make([]byte, pending.packet.Size()+2)
+			binary.BigEndian.PutUint16(data[:2], pending.seq)
+			pending.packet.Serialize(data[2:], 0)
+
+			// Update pending info
+			pending.attempts++
+			pending.sentTime = time.Now()
+			pending.nextRetry = time.Now().Add(s.rto * time.Duration(pending.attempts))
+
+			// Check max retries
+			if pending.attempts > s.maxRetries {
+				s.mu.Lock()
+				delete(s.pendingAcks, pending.seq)
+				s.mu.Unlock()
+				// Connection likely dead
+				s.handleTimeout()
+				return fmt.Errorf("max retransmissions exceeded")
+			}
+
+			atomic.AddUint64(&s.retransmits, 1)
+			s.sendRaw(opcodes.OP_Packet, data)
+
+			s.mu.Lock()
+		} else {
+			// Not time yet, re-queue
+			s.resendQueue = append(s.resendQueue, pending)
+		}
+	}
+
+	// Process reliable queue
+	for len(s.reliableQueue) > 0 {
+		proto := s.reliableQueue[0]
+		s.reliableQueue = s.reliableQueue[1:]
+		s.mu.Unlock()
+
+		if err := s.sendReliable(proto); err != nil {
+			return err
+		}
+
+		s.mu.Lock()
+	}
+
+	// Process unreliable queue
+	for len(s.unreliableQueue) > 0 {
+		proto := s.unreliableQueue[0]
+		s.unreliableQueue = s.unreliableQueue[1:]
+		s.mu.Unlock()
+
+		// Send unreliable (no sequence/retransmit)
+		data := make([]byte, proto.Size())
+		proto.Serialize(data, 0)
+		s.sendRaw(proto.Opcode, data)
+
+		s.mu.Lock()
+	}
+
+	// Check for retransmissions needed
+	now := time.Now()
+	for _, pending := range s.pendingAcks {
+		if now.After(pending.nextRetry) {
+			s.resendQueue = append(s.resendQueue, pending)
+		}
+	}
+
+	s.mu.Unlock()
+
+	// Process any pending ACKs
+	s.sendPendingAcks()
+
+	return nil
+}
+
+// Helper methods
+
+func (s *Stream) processPacketData(data []byte) error {
+	// Create ProtoPacket from raw data
+	proto := packets.NewProtoPacketFromRaw(data, -1, s.opcodeManager)
+
+	// Decompress if needed
+	if err := proto.DecompressPacket(); err != nil {
+		return err
+	}
+
+	// Convert to AppPacket
+	app := proto.MakeApplicationPacket(s.opcodeSize)
+
+	// Deliver to callback
+	if s.onPacket != nil {
+		go s.onPacket(app)
+	}
+
+	return nil
+}
+
+func (s *Stream) scheduleAck() {
+	if s.ackTimer == nil {
+		s.ackTimer = time.AfterFunc(s.ackThreshold, func() {
+			s.sendPendingAcks()
+		})
+	}
+}
+
+func (s *Stream) sendPendingAcks() {
+	s.mu.Lock()
+	if len(s.ackQueue) == 0 {
+		s.mu.Unlock()
+		return
+	}
+
+	// Send all pending ACKs
+	for _, seq := range s.ackQueue {
+		data := make([]byte, 2)
+		binary.BigEndian.PutUint16(data, seq)
+		s.sendRaw(opcodes.OP_Ack, data)
+	}
+
+	s.ackQueue = s.ackQueue[:0]
+	s.lastAckSent = time.Now()
+	s.mu.Unlock()
+}
+
+func (s *Stream) sendAckImmediate(seq uint16) error {
+	data := make([]byte, 2)
+	binary.BigEndian.PutUint16(data, seq)
+	return s.sendRaw(opcodes.OP_Ack, data)
+}
+
+func (s *Stream) sendOutOfOrderAck(seq uint16) error {
+	data := make([]byte, 2)
+	binary.BigEndian.PutUint16(data, seq)
+	return s.sendRaw(opcodes.OP_OutOfOrderAck, data)
+}
+
+func (s *Stream) sendKeepalive() {
+	if s.state.Load() == StateEstablished {
+		s.sendRaw(opcodes.OP_KeepAlive, nil)
+	}
+	// Reschedule
+	if s.keepAliveTimer != nil {
+		s.keepAliveTimer.Reset(10 * time.Second)
+	}
+}
+
+func (s *Stream) resetTimeout() {
+	if s.timeoutTimer != nil {
+		s.timeoutTimer.Reset(30 * time.Second)
+	}
+}
+
+func (s *Stream) handleTimeout() {
+	s.state.Store(StateClosed)
+	if s.onDisconnect != nil {
+		s.onDisconnect()
+	}
+}
+
+func (s *Stream) handleDisconnect() error {
+	s.state.Store(StateClosed)
+	if s.onDisconnect != nil {
+		go s.onDisconnect()
+	}
+	return nil
+}
+
 // sendRaw sends raw data with protocol opcode
 func (s *Stream) sendRaw(opcode uint16, data []byte) error {
 	// Build packet: opcode + data + CRC
@@ -386,121 +849,96 @@ func (s *Stream) sendRaw(opcode uint16, data []byte) error {
 	return s.conn.AsyncWrite(packet, nil)
 }
 
-// Helper methods
-
-func (s *Stream) processPacketData(data []byte) error {
-	// Create ProtoPacket from raw data
-	proto := packets.NewProtoPacketFromRaw(data, -1, s.opcodeManager)
-
-	// Decompress if needed
-	proto.DecompressPacket()
-
-	// Convert to AppPacket
-	app := proto.MakeApplicationPacket(s.opcodeSize)
-
-	// Deliver to callback
-	if s.onPacket != nil {
-		s.onPacket(app)
-	}
-
-	return nil
-}
-
-func (s *Stream) scheduleAck(seq uint16) {
-	if s.ackTimer == nil {
-		s.ackTimer = time.AfterFunc(s.ackThreshold, func() {
-			s.sendAckImmediate(seq)
-		})
-	}
-}
-
-func (s *Stream) sendAckImmediate(seq uint16) error {
-	data := make([]byte, 2)
-	binary.BigEndian.PutUint16(data, seq)
-	return s.sendRaw(opcodes.OP_Ack, data)
-}
-
-func (s *Stream) sendKeepalive() {
-	s.sendRaw(opcodes.OP_KeepAlive, nil)
-	// Reschedule
-	if s.keepAliveTimer != nil {
-		s.keepAliveTimer.Reset(10 * time.Second)
-	}
-}
-
-func (s *Stream) resetTimeout() {
-	if s.timeoutTimer != nil {
-		s.timeoutTimer.Reset(30 * time.Second)
-	}
-}
-
 func (s *Stream) appToProto(app *packets.AppPacket) *packets.ProtoPacket {
 	proto := packets.NewProtoPacket(app.Opcode, app.Buffer, s.opcodeManager)
 	proto.CopyInfo(app.Packet)
+	proto.CompressThreshold = 100
+	proto.EncodeKey = s.encodeKey
 	return proto
 }
 
-func (s *Stream) processQueues() error {
-	// Process reliable queue first
-	s.mu.Lock()
-	for len(s.reliableQueue) > 0 {
-		proto := s.reliableQueue[0]
-		s.reliableQueue = s.reliableQueue[1:]
-		s.mu.Unlock()
-		if err := s.sendReliable(proto); err != nil {
-			return err
-		}
-		s.mu.Lock()
-	}
-	s.mu.Unlock()
-
-	return nil
-}
-
-func (s *Stream) sendFragmented(proto *packets.ProtoPacket) error {
-	// Fragment implementation
-	return nil // TODO
-}
-
-func (s *Stream) handleSessionRequest(data []byte) error {
-	// Session setup
-	s.state.Store(StateConnecting)
-	return nil
-}
-
-func (s *Stream) handleSessionResponse(data []byte) error {
-	s.state.Store(StateEstablished)
-	return nil
-}
-
-func (s *Stream) handleCombined(data []byte) error {
-	// Process combined packets
-	return nil
-}
-
-func (s *Stream) handleAppCombined(data []byte) error {
-	// Process app-combined packets
-	return nil
-}
-
-func (s *Stream) handleDisconnect() error {
-	s.state.Store(StateClosed)
-	if s.onDisconnect != nil {
-		s.onDisconnect()
-	}
-	return nil
-}
+// Public methods
 
 // SetPacketCallback sets the callback for received packets
 func (s *Stream) SetPacketCallback(fn func(*packets.AppPacket)) {
 	s.onPacket = fn
 }
 
+// SetDisconnectCallback sets the callback for disconnection
+func (s *Stream) SetDisconnectCallback(fn func()) {
+	s.onDisconnect = fn
+}
+
+// GetState returns current stream state
+func (s *Stream) GetState() StreamState {
+	return s.state.Load().(StreamState)
+}
+
+// IsConnected returns true if stream is established
+func (s *Stream) IsConnected() bool {
+	return s.GetState() == StateEstablished
+}
+
+// SendSessionRequest initiates client connection
+func (s *Stream) SendSessionRequest() error {
+	// Build session request packet
+	// Format: version(4) + sessionID(4) + maxLen(2)
+	data := make([]byte, 10)
+	binary.BigEndian.PutUint32(data[0:4], 2) // Protocol version
+	binary.BigEndian.PutUint32(data[4:8], s.sessionID)
+	binary.BigEndian.PutUint16(data[8:10], s.maxLen)
+
+	s.state.Store(StateConnecting)
+	return s.sendRaw(opcodes.OP_SessionRequest, data)
+}
+
+// validateSession checks if packet belongs to this session
+func (s *Stream) validateSession(sessionID uint32) bool {
+	if s.sessionID == 0 {
+		// Not yet established, accept any
+		return true
+	}
+	if sessionID != s.sessionID {
+		// Wrong session - send out of session response
+		go s.sendRaw(opcodes.OP_OutOfSession, nil)
+		return false
+	}
+	return true
+}
+
+// GetSessionID returns current session ID
+func (s *Stream) GetSessionID() uint32 {
+	s.mu.RLock()
+	defer s.mu.RUnlock()
+	return s.sessionID
+}
+
+// SetSessionID sets the session ID (for client mode)
+func (s *Stream) SetSessionID(id uint32) {
+	s.mu.Lock()
+	s.sessionID = id
+	s.mu.Unlock()
+}
+
+// GetRemoteAddr returns remote address
+func (s *Stream) GetRemoteAddr() string {
+	if s.conn != nil {
+		return s.conn.RemoteAddr().String()
+	}
+	return ""
+}
+
 // Close closes the stream
 func (s *Stream) Close() error {
+	if s.state.Load() == StateClosed {
+		return nil
+	}
+
 	s.state.Store(StateClosed)
+
 	// Send disconnect
 	s.sendRaw(opcodes.OP_SessionDisconnect, nil)
+
 	// Clean up timers
 	if s.keepAliveTimer != nil {
 		s.keepAliveTimer.Stop()
@@ -508,5 +946,28 @@ func (s *Stream) Close() error {
 	if s.ackTimer != nil {
 		s.ackTimer.Stop()
 	}
+	if s.timeoutTimer != nil {
+		s.timeoutTimer.Stop()
+	}
+
+	// Clear queues
+	s.mu.Lock()
+	s.reliableQueue = nil
+	s.unreliableQueue = nil
+	s.resendQueue = nil
+	s.pendingAcks = nil
+	s.fragments = nil
+	s.outOfOrder = nil
+	s.mu.Unlock()
+
 	return nil
 }
+
+// Utility functions
+
+func absTime(d time.Duration) time.Duration {
+	if d < 0 {
+		return -d
+	}
+	return d
+}