sync with private branch

16 files changed, 225 insertions, 101 deletions

diff --git a/cryptest.dsp b/cryptest.dsp
index b7b826ff..31c21ba3 100644
--- a/cryptest.dsp
+++ b/cryptest.dsp
@@ -53,7 +53,8 @@ BSC32=bscmake.exe
 # ADD BSC32 /nologo
 LINK32=link.exe
 # ADD BASE LINK32 kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib Ws2_32.lib /nologo /subsystem:console /debug /machine:I386 /OPT:NOWIN98
-# ADD LINK32 kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib Ws2_32.lib /nologo /subsystem:console /debug /machine:I386 /OPT:NOWIN98
+# ADD LINK32 kernel32.lib user32.lib gdi32.lib winspool.lib comdlg32.lib advapi32.lib shell32.lib ole32.lib oleaut32.lib uuid.lib odbc32.lib odbccp32.lib Ws2_32.lib /nologo /subsystem:console /debug /machine:I386 /OPT:NOWIN98 /OPT:REF /OPT:ICF
+# SUBTRACT LINK32 /pdb:none
 # Begin Special Build Tool
 SOURCE="$(InputPath)"
 PostBuild_Cmds=echo This configuration is used to build a static binary for FIPS 140 evaluation by a testing laboratory.	echo Crypto++ users should not build this configuration directly.
diff --git a/filters.cpp b/filters.cpp
index 47b5c93e..576022e3 100644
--- a/filters.cpp
+++ b/filters.cpp
@@ -98,6 +98,15 @@ void Filter::PropagateInitialize(const NameValuePairs &parameters, int propagati
 		AttachedTransformation()->ChannelInitialize(channel, parameters, propagation-1);
 }
 
+unsigned int Filter::OutputModifiable(int outputSite, byte *inString, unsigned int length, int messageEnd, bool blocking, const std::string &channel)
+{
+	if (messageEnd)
+		messageEnd--;
+	unsigned int result = AttachedTransformation()->PutModifiable2(inString, length, messageEnd, blocking);
+	m_continueAt = result ? outputSite : 0;
+	return result;
+}
+
 unsigned int Filter::Output(int outputSite, const byte *inString, unsigned int length, int messageEnd, bool blocking, const std::string &channel)
 {
 	if (messageEnd)
@@ -152,6 +161,27 @@ unsigned int MeterFilter::Put2(const byte *begin, unsigned int length, int messa
 	return 0;
 }
 
+unsigned int MeterFilter::PutModifiable2(byte *begin, unsigned int length, int messageEnd, bool blocking)
+{
+	if (m_transparent)
+	{
+		FILTER_BEGIN;
+		m_currentMessageBytes += length;
+		m_totalBytes += length;
+
+		if (messageEnd)
+		{
+			m_currentMessageBytes = 0;
+			m_currentSeriesMessages++;
+			m_totalMessages++;
+		}
+		
+		FILTER_OUTPUT_MODIFIABLE(1, begin, length, messageEnd);
+		FILTER_END_NO_MESSAGE_END;
+	}
+	return 0;
+}
+
 bool MeterFilter::IsolatedMessageSeriesEnd(bool blocking)
 {
 	m_currentMessageBytes = 0;
@@ -414,12 +444,18 @@ void ProxyFilter::SetFilter(Filter *filter)
 	}
 }
 
-void ProxyFilter::NextPutMultiple(const byte *s, unsigned int len) 
+void ProxyFilter::NextPutMultiple(const byte *s, unsigned int len)
 {
 	if (m_filter.get())
 		m_filter->Put(s, len);
 }
 
+void ProxyFilter::NextPutModifiable(byte *s, unsigned int len)
+{
+	if (m_filter.get())
+		m_filter->PutModifiable(s, len);
+}
+
 // *************************************************************
 
 unsigned int ArraySink::Put2(const byte *begin, unsigned int length, int messageEnd, bool blocking)
diff --git a/filters.h b/filters.h
index 22994c95..cc33e28d 100644
--- a/filters.h
+++ b/filters.h
@@ -39,6 +39,7 @@ protected:
 	void PropagateInitialize(const NameValuePairs &parameters, int propagation, const std::string &channel=NULL_CHANNEL);
 
 	unsigned int Output(int outputSite, const byte *inString, unsigned int length, int messageEnd, bool blocking, const std::string &channel=NULL_CHANNEL);
+	unsigned int OutputModifiable(int outputSite, byte *inString, unsigned int length, int messageEnd, bool blocking, const std::string &channel=NULL_CHANNEL);
 	bool OutputMessageEnd(int outputSite, int propagation, bool blocking, const std::string &channel=NULL_CHANNEL);
 	bool OutputFlush(int outputSite, bool hardFlush, int propagation, bool blocking, const std::string &channel=NULL_CHANNEL);
 	bool OutputMessageSeriesEnd(int outputSite, int propagation, bool blocking, const std::string &channel=NULL_CHANNEL);
@@ -94,7 +95,10 @@ public:
 	unsigned int GetTotalMessages() {return m_totalMessages;}
 	unsigned int GetTotalMessageSeries() {return m_totalMessageSeries;}
 
+	byte * CreatePutSpace(unsigned int &size)
+		{return AttachedTransformation()->CreatePutSpace(size);}
 	unsigned int Put2(const byte *begin, unsigned int length, int messageEnd, bool blocking);
+	unsigned int PutModifiable2(byte *inString, unsigned int length, int messageEnd, bool blocking);
 	bool IsolatedMessageSeriesEnd(bool blocking);
 
 private:
@@ -388,16 +392,23 @@ public:
 	bool GetPassWaitObjects() const {return (m_behavior & PASS_WAIT_OBJECTS) != 0;}
 	void SetPassWaitObjects(bool pass) { if (pass) m_behavior |= PASS_WAIT_OBJECTS; else m_behavior &= ~(word32) PASS_WAIT_OBJECTS; }
 
-	unsigned int Put2(const byte *begin, unsigned int length, int messageEnd, bool blocking)
-		{return m_target ? m_target->Put2(begin, length, GetPassSignals() ? messageEnd : 0, blocking) : 0;}
+	bool CanModifyInput() const
+		{return m_target ? m_target->CanModifyInput() : false;}
+
 	void Initialize(const NameValuePairs &parameters, int propagation)
 		{ChannelInitialize(NULL_CHANNEL, parameters, propagation);}
+	byte * CreatePutSpace(unsigned int &size)
+		{return m_target ? m_target->CreatePutSpace(size) : (size=0, NULL);}
+	unsigned int Put2(const byte *begin, unsigned int length, int messageEnd, bool blocking)
+		{return m_target ? m_target->Put2(begin, length, GetPassSignals() ? messageEnd : 0, blocking) : 0;}
 	bool Flush(bool hardFlush, int propagation=-1, bool blocking=true)
 		{return m_target && GetPassSignals() ? m_target->Flush(hardFlush, propagation, blocking) : false;}
 	bool MessageSeriesEnd(int propagation=-1, bool blocking=true)
 		{return m_target && GetPassSignals() ? m_target->MessageSeriesEnd(propagation, blocking) : false;}
 
 	void ChannelInitialize(const std::string &channel, const NameValuePairs &parameters=g_nullNameValuePairs, int propagation=-1);
+	byte * ChannelCreatePutSpace(const std::string &channel, unsigned int &size)
+		{return m_target ? m_target->ChannelCreatePutSpace(channel, size) : (size=0, NULL);}
 	unsigned int ChannelPut2(const std::string &channel, const byte *begin, unsigned int length, int messageEnd, bool blocking)
 		{return m_target ? m_target->ChannelPut2(channel, begin, length, GetPassSignals() ? messageEnd : 0, blocking) : 0;}
 	unsigned int ChannelPutModifiable2(const std::string &channel, byte *begin, unsigned int length, int messageEnd, bool blocking)
@@ -426,6 +437,8 @@ public:
 	bool GetPassSignal() const {return m_passSignal;}
 	void SetPassSignal(bool passSignal) {m_passSignal = passSignal;}
 
+	byte * CreatePutSpace(unsigned int &size)
+		{return m_owner.AttachedTransformation()->CreatePutSpace(size);}
 	unsigned int Put2(const byte *begin, unsigned int length, int messageEnd, bool blocking)
 		{return m_owner.AttachedTransformation()->Put2(begin, length, m_passSignal ? messageEnd : 0, blocking);}
 	unsigned int PutModifiable2(byte *begin, unsigned int length, int messageEnd, bool blocking)
@@ -463,6 +476,7 @@ public:
 
 	void SetFilter(Filter *filter);
 	void NextPutMultiple(const byte *s, unsigned int len);
+	void NextPutModifiable(byte *inString, unsigned int length);
 
 protected:
 	member_ptr<BufferedTransformation> m_filter;
diff --git a/fltrimpl.h b/fltrimpl.h
index a35e68b0..60c827be 100644
--- a/fltrimpl.h
+++ b/fltrimpl.h
@@ -39,4 +39,15 @@
 #define FILTER_OUTPUT_BYTE(site, output)	\
 	FILTER_OUTPUT(site, &(const byte &)(byte)output, 1, 0)
 
+#define FILTER_OUTPUT2_MODIFIABLE(site, statement, output, length, messageEnd)	\
+	{\
+	case site:	\
+	statement;	\
+	if (OutputModifiable(site, output, length, messageEnd, blocking))	\
+		return STDMAX(1U, (unsigned int)length-m_inputPosition);\
+	}
+
+#define FILTER_OUTPUT_MODIFIABLE(site, output, length, messageEnd)	\
+	FILTER_OUTPUT2_MODIFIABLE(site, 0, output, length, messageEnd)
+
 #endif
diff --git a/integer.cpp b/integer.cpp
index 7e8e3f8c..260f6f0e 100644
--- a/integer.cpp
+++ b/integer.cpp
@@ -16,6 +16,8 @@
 
 #ifdef SSE2_INTRINSICS_AVAILABLE
 #include <emmintrin.h>
+#elif defined(_MSC_VER) && defined(_M_IX86)
+#pragma message("You do no seem to have the Visual C++ Processor Pack installed, so use of SSE2 intrinsics will be disabled.")
 #endif
 
 #include "algebra.cpp"
@@ -31,27 +33,30 @@ bool FunctionAssignIntToInteger(const std::type_info &valueType, void *pInteger,
 	return true;
 }
 
-static int DummyAssignIntToInteger = (AssignIntToInteger = FunctionAssignIntToInteger, 0);
+static const char s_RunAtStartup = (AssignIntToInteger = FunctionAssignIntToInteger, 0);
 
-#ifdef SSE2_INTRINSICS_AVAILABLE
+#if defined(SSE2_INTRINSICS_AVAILABLE) || defined(_MSC_VER)
 template <class T>
 CPP_TYPENAME AllocatorBase<T>::pointer AlignedAllocator<T>::allocate(size_type n, const void *)
 {
-	if (n < 4)
-		return new T[n];
-	else
+#ifdef SSE2_INTRINSICS_AVAILABLE
+	if (n >= 4)
 		return (T *)_mm_malloc(sizeof(T)*n, 16);
-
+	else
+#endif
+		return new T[n];
 }
 
 template <class T>
 void AlignedAllocator<T>::deallocate(void *p, size_type n)
 {
 	memset(p, 0, n*sizeof(T));
-	if (n < 4)
-		delete [] p;
-	else
+#ifdef SSE2_INTRINSICS_AVAILABLE
+	if (n >= 4)
 		_mm_free(p);
+	else
+#endif
+		delete [] p;
 }
 
 template class AlignedAllocator<word>;
diff --git a/integer.h b/integer.h
index fcf3ebef..503a74f9 100644
--- a/integer.h
+++ b/integer.h
@@ -23,7 +23,7 @@
 
 NAMESPACE_BEGIN(CryptoPP)
 
-#ifdef SSE2_INTRINSICS_AVAILABLE
+#if defined(SSE2_INTRINSICS_AVAILABLE) || defined(_MSC_VER)
 	template <class T>
 	class AlignedAllocator : public AllocatorBase<T>
 	{
diff --git a/network.cpp b/network.cpp
index 72001d4e..f7705a70 100644
--- a/network.cpp
+++ b/network.cpp
@@ -2,6 +2,7 @@
 
 #include "pch.h"
 #include "network.h"
+#include "wait.h"
 
 NAMESPACE_BEGIN(CryptoPP)
 
@@ -33,10 +34,24 @@ bool NonblockingSink::IsolatedFlush(bool hardFlush, bool blocking)
 #ifdef HIGHRES_TIMER_AVAILABLE
 
 NetworkSource::NetworkSource(BufferedTransformation *attachment)
-	: NonblockingSource(attachment), m_buf(1024*4), m_bufSize(0), m_state(NORMAL)
+	: NonblockingSource(attachment), m_buf(1024*16)
+	, m_waitingForResult(false), m_outputBlocked(false)
+	, m_dataBegin(0), m_dataEnd(0)
 {
 }
 
+void NetworkSource::GetWaitObjects(WaitObjectContainer &container)
+{
+	if (!m_outputBlocked)
+	{
+		if (m_dataBegin == m_dataEnd)
+			AccessReceiver().GetWaitObjects(container); 
+		else
+			container.SetNoWait();
+	}
+	AttachedTransformation()->GetWaitObjects(container);
+}
+
 unsigned int NetworkSource::GeneralPump2(unsigned long &byteCount, bool blockingOutput, unsigned long maxTime, bool checkDelimiter, byte delimiter)
 {
 	NetworkReceiver &receiver = AccessReceiver();
@@ -45,80 +60,93 @@ unsigned int NetworkSource::GeneralPump2(unsigned long &byteCount, bool blocking
 	byteCount = 0;
 	bool forever = maxTime == INFINITE_TIME;
 	Timer timer(Timer::MILLISECONDS, forever);
-	unsigned long timeout;
 	BufferedTransformation *t = AttachedTransformation();
 
-	if (m_state == OUTPUT_BLOCKED)
+	if (m_outputBlocked)
 		goto DoOutput;
 
 	while (true)
 	{
-		if (m_state == WAITING_FOR_RESULT)
+		if (m_dataBegin == m_dataEnd)
 		{
-			if (receiver.MustWaitForResult())
+			if (receiver.EofReceived())
+				break;
+
+			if (m_waitingForResult)
 			{
-				timeout = SaturatingSubtract(maxTime, timer.ElapsedTime());
-				if (!receiver.Wait(timeout))
+				if (receiver.MustWaitForResult() && !receiver.Wait(SaturatingSubtract(maxTime, timer.ElapsedTime())))
 					break;
+
+				unsigned int recvResult = receiver.GetReceiveResult();
+				m_dataEnd += recvResult;
+				m_waitingForResult = false;
+
+				if (!receiver.MustWaitToReceive() && !receiver.EofReceived() && m_dataEnd != m_buf.size())
+					goto ReceiveNoWait;
 			}
+			else
+			{
+				m_dataEnd = m_dataBegin = 0;
 
-			unsigned int recvResult = receiver.GetReceiveResult();
-//			assert(recvResult > 0 || receiver.EofReceived());
-			m_bufSize += recvResult;
-			m_state = NORMAL;
-		}
+				if (receiver.MustWaitToReceive())
+				{
+					if (!receiver.Wait(SaturatingSubtract(maxTime, timer.ElapsedTime())))
+						break;
 
-		if (m_bufSize == 0)
-		{
-			if (receiver.EofReceived())
-				break;
+					receiver.Receive(m_buf+m_dataEnd, m_buf.size()-m_dataEnd);
+					m_waitingForResult = true;
+				}
+				else
+				{
+ReceiveNoWait:
+					m_waitingForResult = true;
+					// call Receive repeatedly as long as data is immediately available,
+					// because some receivers tend to return data in small pieces
+					while (receiver.Receive(m_buf+m_dataEnd, m_buf.size()-m_dataEnd))
+					{
+						unsigned int recvResult = receiver.GetReceiveResult();
+						m_dataEnd += recvResult;
+						if (receiver.EofReceived() || m_dataEnd == m_buf.size())
+						{
+							m_waitingForResult = false;
+							break;
+						}
+					}
+				}
+			}
 		}
 		else
 		{
-			m_putSize = STDMIN((unsigned long)m_bufSize, maxSize - byteCount);
+			m_putSize = STDMIN((unsigned long)m_dataEnd-m_dataBegin, maxSize-byteCount);
 			if (checkDelimiter)
-				m_putSize = std::find(m_buf.begin(), m_buf+m_putSize, delimiter) - m_buf;
+				m_putSize = std::find(m_buf+m_dataBegin, m_buf+m_dataBegin+m_putSize, delimiter) - (m_buf+m_dataBegin);
 
 DoOutput:
-			unsigned int result = t->PutModifiable2(m_buf, m_putSize, 0, forever || blockingOutput);
+			unsigned int result = t->PutModifiable2(m_buf+m_dataBegin, m_putSize, 0, forever || blockingOutput);
 			if (result)
 			{
-				timeout = SaturatingSubtract(maxTime, timer.ElapsedTime());
-				if (t->Wait(timeout))
+				if (t->Wait(SaturatingSubtract(maxTime, timer.ElapsedTime())))
 					goto DoOutput;
 				else
 				{
-					m_state = OUTPUT_BLOCKED;
+					m_outputBlocked = true;
 					return result;
 				}
 			}
-			m_state = NORMAL;
+			m_outputBlocked = false;
 
 			byteCount += m_putSize;
-			m_bufSize -= m_putSize;
-			if (m_bufSize > 0)
-			{
-				memmove(m_buf, m_buf+m_putSize, m_bufSize);
-				if (checkDelimiter && m_buf[0] == delimiter)
-					break;
-			}
-		}
-
-		if (byteCount == maxSize)
-			break;
-
-		unsigned long elapsed = timer.ElapsedTime();
-		if (elapsed > maxTime)
-			break;	// once time limit is reached, return even if there is more data waiting
-
-		if (receiver.MustWaitToReceive())
-		{
-			if (!receiver.Wait(maxTime - elapsed))
+			m_dataBegin += m_putSize;
+			if (checkDelimiter && m_dataBegin < m_dataEnd && m_buf[m_dataBegin] == delimiter)
+				break;
+			if (byteCount == maxSize)
+				break;
+			// once time limit is reached, return even if there is more data waiting
+			// but make 0 a special case so caller can request a large amount of data to be
+			// pumped as long as it is immediately available
+			if (maxTime > 0 && timer.ElapsedTime() > maxTime)
 				break;
 		}
-
-		receiver.Receive(m_buf+m_bufSize, m_buf.size()-m_bufSize);
-		m_state = WAITING_FOR_RESULT;
 	}
 
 	return 0;
@@ -126,6 +154,12 @@ DoOutput:
 
 // *************************************************************
 
+NetworkSink::NetworkSink(unsigned int maxBufferSize, bool autoFlush)
+	: m_maxBufferSize(maxBufferSize), m_autoFlush(autoFlush)
+	, m_needSendResult(false), m_buffer(STDMIN(16U*1024U, maxBufferSize)), m_blockedBytes(0) 
+{
+}
+
 unsigned int NetworkSink::Put2(const byte *inString, unsigned int length, int messageEnd, bool blocking)
 {
 	if (m_blockedBytes)
@@ -134,7 +168,7 @@ unsigned int NetworkSink::Put2(const byte *inString, unsigned int length, int me
 		inString += length - m_blockedBytes;
 		length = m_blockedBytes;
 	}
-	m_buffer.LazyPut(inString, length);
+	LazyPutter lp(m_buffer, inString, length);
 
 	unsigned int targetSize = messageEnd ? 0 : m_maxBufferSize;
 	TimedFlush(blocking ? INFINITE_TIME : 0, m_autoFlush ? 0 : targetSize);
@@ -144,7 +178,6 @@ unsigned int NetworkSink::Put2(const byte *inString, unsigned int length, int me
 		assert(!blocking);
 		m_blockedBytes = STDMIN(m_buffer.CurrentSize() - targetSize, (unsigned long)length);
 		m_buffer.UndoLazyPut(m_blockedBytes);
-		m_buffer.FinalizeLazyPut();
 		return STDMAX(m_blockedBytes, 1U);
 	}
 	m_blockedBytes = 0;
@@ -156,51 +189,43 @@ unsigned int NetworkSink::Put2(const byte *inString, unsigned int length, int me
 
 unsigned int NetworkSink::TimedFlush(unsigned long maxTime, unsigned int targetSize)
 {
-	if (m_buffer.IsEmpty())
-		return 0;
-
 	NetworkSender &sender = AccessSender();
 
 	bool forever = maxTime == INFINITE_TIME;
 	Timer timer(Timer::MILLISECONDS, forever);
-	unsigned long timeout;
 	unsigned int totalFlushSize = 0;
 
 	while (true)
 	{
+		if (m_buffer.CurrentSize() <= targetSize)
+			break;
+		
 		if (m_needSendResult)
 		{
-			if (sender.MustWaitForResult())
-			{
-				timeout = SaturatingSubtract(maxTime, timer.ElapsedTime());
-				if (!sender.Wait(timeout))
-					break;
-			}
+			if (sender.MustWaitForResult() && !sender.Wait(SaturatingSubtract(maxTime, timer.ElapsedTime())))
+				break;
 
 			unsigned int sendResult = sender.GetSendResult();
 			m_buffer.Skip(sendResult);
 			totalFlushSize += sendResult;
 			m_needSendResult = false;
 
-			if (m_buffer.CurrentSize() <= targetSize)
+			if (!m_buffer.AnyRetrievable())
 				break;
 		}
 
-		unsigned long elapsed = timer.ElapsedTime();
-		if (elapsed > maxTime)
-			break;	// once time limit is reached, return even if there is more data waiting
-
-		if (sender.MustWaitToSend())
-		{
-			if (!sender.Wait(maxTime - elapsed))
-				break;
-		}
+		unsigned long timeOut = maxTime ? SaturatingSubtract(maxTime, timer.ElapsedTime()) : 0;
+		if (sender.MustWaitToSend() && !sender.Wait(timeOut))
+			break;
 
 		unsigned int contiguousSize = 0;
 		const byte *block = m_buffer.Spy(contiguousSize);
 
 		sender.Send(block, contiguousSize);
 		m_needSendResult = true;
+
+		if (maxTime > 0 && timeOut == 0)
+			break;	// once time limit is reached, return even if there is more data waiting
 	}
 
 	return totalFlushSize;
diff --git a/network.h b/network.h
index e2ec59a6..790a00e1 100644
--- a/network.h
+++ b/network.h
@@ -45,7 +45,8 @@ class CRYPTOPP_NO_VTABLE NetworkReceiver : public Waitable
 public:
 	virtual bool MustWaitToReceive() {return false;}
 	virtual bool MustWaitForResult() {return false;}
-	virtual void Receive(byte* buf, unsigned int bufLen) =0;
+	//! receive data from network source, returns whether result is immediately available
+	virtual bool Receive(byte* buf, unsigned int bufLen) =0;
 	virtual unsigned int GetReceiveResult() =0;
 	virtual bool EofReceived() const =0;
 };
@@ -96,8 +97,7 @@ public:
 
 	unsigned int GetMaxWaitObjectCount() const
 		{return GetReceiver().GetMaxWaitObjectCount() + AttachedTransformation()->GetMaxWaitObjectCount();}
-	void GetWaitObjects(WaitObjectContainer &container)
-		{AccessReceiver().GetWaitObjects(container); AttachedTransformation()->GetWaitObjects(container);}
+	void GetWaitObjects(WaitObjectContainer &container);
 
 	unsigned int GeneralPump2(unsigned long &byteCount, bool blockingOutput=true, unsigned long maxTime=INFINITE_TIME, bool checkDelimiter=false, byte delimiter='\n');
 	bool SourceExhausted() const {return GetReceiver().EofReceived();}
@@ -107,17 +107,16 @@ protected:
 	const NetworkReceiver & GetReceiver() const {return const_cast<NetworkSource *>(this)->AccessReceiver();}
 
 private:
-	enum {NORMAL, WAITING_FOR_RESULT, OUTPUT_BLOCKED};
 	SecByteBlock m_buf;
-	unsigned int m_bufSize, m_putSize, m_state;
+	unsigned int m_putSize, m_dataBegin, m_dataEnd;
+	bool m_waitingForResult, m_outputBlocked;
 };
 
 //! Network Sink
 class CRYPTOPP_NO_VTABLE NetworkSink : public NonblockingSink
 {
 public:
-	NetworkSink(unsigned int maxBufferSize, bool autoFlush)
-		: m_maxBufferSize(maxBufferSize), m_autoFlush(autoFlush), m_needSendResult(false), m_blockedBytes(0) {}
+	NetworkSink(unsigned int maxBufferSize, bool autoFlush);
 
 	unsigned int GetMaxWaitObjectCount() const
 		{return GetSender().GetMaxWaitObjectCount();}
diff --git a/queue.cpp b/queue.cpp
index b5ac16de..976bb0c5 100644
--- a/queue.cpp
+++ b/queue.cpp
@@ -234,8 +234,18 @@ void ByteQueue::LazyPut(const byte *inString, unsigned int size)
 {
 	if (m_lazyLength > 0)
 		FinalizeLazyPut();
+	m_lazyString = const_cast<byte *>(inString);
+	m_lazyLength = size;
+	m_lazyStringModifiable = false;
+}
+
+void ByteQueue::LazyPutModifiable(byte *inString, unsigned int size)
+{
+	if (m_lazyLength > 0)
+		FinalizeLazyPut();
 	m_lazyString = inString;
 	m_lazyLength = size;
+	m_lazyStringModifiable = true;
 }
 
 void ByteQueue::UndoLazyPut(unsigned int size)
@@ -309,7 +319,10 @@ unsigned int ByteQueue::TransferTo2(BufferedTransformation &target, unsigned lon
 		unsigned int len = (unsigned int)STDMIN(bytesLeft, (unsigned long)m_lazyLength);
 		if (len)
 		{
-			target.ChannelPut(channel, m_lazyString, len);
+			if (m_lazyStringModifiable)
+				target.ChannelPutModifiable(channel, m_lazyString, len);
+			else
+				target.ChannelPut(channel, m_lazyString, len);
 			m_lazyString += len;
 			m_lazyLength -= len;
 			bytesLeft -= len;
diff --git a/queue.h b/queue.h
index e247d9f6..b9641bc4 100644
--- a/queue.h
+++ b/queue.h
@@ -52,6 +52,7 @@ public:
 	const byte * Spy(unsigned int &contiguousSize) const;
 
 	void LazyPut(const byte *inString, unsigned int size);
+	void LazyPutModifiable(byte *inString, unsigned int size);
 	void UndoLazyPut(unsigned int size);
 	void FinalizeLazyPut();
 
@@ -100,8 +101,9 @@ private:
 
 	unsigned int m_nodeSize;
 	ByteQueueNode *m_head, *m_tail;
-	const byte *m_lazyString;
+	byte *m_lazyString;
 	unsigned int m_lazyLength;
+	bool m_lazyStringModifiable;
 };
 
 //! use this to make sure LazyPut is finalized in event of exception
@@ -112,10 +114,20 @@ public:
 		: m_bq(bq) {bq.LazyPut(inString, size);}
 	~LazyPutter()
 		{try {m_bq.FinalizeLazyPut();} catch(...) {}}
+protected:
+	LazyPutter(ByteQueue &bq) : m_bq(bq) {}
 private:
 	ByteQueue &m_bq;
 };
 
+//! like LazyPutter, but does a LazyPutModifiable instead
+class LazyPutterModifiable : public LazyPutter
+{
+public:
+	LazyPutterModifiable(ByteQueue &bq, byte *inString, unsigned int size)
+		: LazyPutter(bq) {bq.LazyPutModifiable(inString, size);}
+};
+
 NAMESPACE_END
 
 NAMESPACE_BEGIN(std)
diff --git a/socketft.cpp b/socketft.cpp
index 0f7b7eb0..d7515939 100644
--- a/socketft.cpp
+++ b/socketft.cpp
@@ -311,12 +311,13 @@ SocketReceiver::SocketReceiver(Socket &s)
 	m_overlapped.hEvent = m_event;
 }
 
-void SocketReceiver::Receive(byte* buf, unsigned int bufLen)
+bool SocketReceiver::Receive(byte* buf, unsigned int bufLen)
 {
 	assert(!m_resultPending && !m_eofReceived);
 
 	DWORD flags = 0;
-	WSABUF wsabuf = {bufLen, (char *)buf};
+	// don't queue too much at once, or we might use up non-paged memory
+	WSABUF wsabuf = {STDMIN(bufLen, 128U*1024U), (char *)buf};
 	if (WSARecv(m_s, &wsabuf, 1, &m_lastResult, &flags, &m_overlapped, NULL) == 0)
 	{
 		if (m_lastResult == 0)
@@ -336,6 +337,7 @@ void SocketReceiver::Receive(byte* buf, unsigned int bufLen)
 			m_resultPending = true;
 		}
 	}
+	return !m_resultPending;
 }
 
 void SocketReceiver::GetWaitObjects(WaitObjectContainer &container)
@@ -386,7 +388,8 @@ SocketSender::SocketSender(Socket &s)
 void SocketSender::Send(const byte* buf, unsigned int bufLen)
 {
 	DWORD written = 0;
-	WSABUF wsabuf = {bufLen, (char *)buf};
+	// don't queue too much at once, or we might use up non-paged memory
+	WSABUF wsabuf = {STDMIN(bufLen, 128U*1024U), (char *)buf};
 	if (WSASend(m_s, &wsabuf, 1, &written, 0, &m_overlapped, NULL) == 0)
 	{
 		m_resultPending = false;
@@ -436,11 +439,12 @@ void SocketReceiver::GetWaitObjects(WaitObjectContainer &container)
 		container.AddReadFd(m_s);
 }
 
-void SocketReceiver::Receive(byte* buf, unsigned int bufLen)
+bool SocketReceiver::Receive(byte* buf, unsigned int bufLen)
 {
 	m_lastResult = m_s.Receive(buf, bufLen);
 	if (bufLen > 0 && m_lastResult == 0)
 		m_eofReceived = true;
+	return true;
 }
 
 unsigned int SocketReceiver::GetReceiveResult()
diff --git a/socketft.h b/socketft.h
index 2ce54549..e8baef0f 100644
--- a/socketft.h
+++ b/socketft.h
@@ -130,7 +130,7 @@ public:
 #else
 	bool MustWaitForResult() {return true;}
 #endif
-	void Receive(byte* buf, unsigned int bufLen);
+	bool Receive(byte* buf, unsigned int bufLen);
 	unsigned int GetReceiveResult();
 	bool EofReceived() const {return m_eofReceived;}
 
diff --git a/test.cpp b/test.cpp
index 63218f20..fafcb2d6 100644
--- a/test.cpp
+++ b/test.cpp
@@ -805,6 +805,7 @@ void ForwardTcpPort(const char *sourcePortName, const char *destinationHost, con
 
 	sockListen.Create();
 	sockListen.Bind(sourcePort);
+	setsockopt(sockListen, IPPROTO_TCP, TCP_NODELAY, "\x01", 1);
 
 	cout << "Listing on port " << sourcePort << ".\n";
 	sockListen.Listen();
diff --git a/wait.h b/wait.h
index c1435221..db9caa6f 100644
--- a/wait.h
+++ b/wait.h
@@ -9,7 +9,7 @@
 #include <vector>
 
 #ifdef USE_WINDOWS_STYLE_SOCKETS
-#include <windows.h>
+#include <winsock2.h>
 #else
 #include <sys/types.h>
 #endif
diff --git a/winpipes.cpp b/winpipes.cpp
index a195d174..d125a6be 100644
--- a/winpipes.cpp
+++ b/winpipes.cpp
@@ -86,12 +86,13 @@ WindowsPipeReceiver::WindowsPipeReceiver()
 	m_overlapped.hEvent = m_event;
 }
 
-void WindowsPipeReceiver::Receive(byte* buf, unsigned int bufLen)
+bool WindowsPipeReceiver::Receive(byte* buf, unsigned int bufLen)
 {
 	assert(!m_resultPending && !m_eofReceived);
 
 	HANDLE h = GetHandle();
-	if (ReadFile(h, buf, bufLen, &m_lastResult, &m_overlapped))
+	// don't queue too much at once, or we might use up non-paged memory
+	if (ReadFile(h, buf, STDMIN(bufLen, 128U*1024U), &m_lastResult, &m_overlapped))
 	{
 		if (m_lastResult == 0)
 			m_eofReceived = true;
@@ -111,6 +112,7 @@ void WindowsPipeReceiver::Receive(byte* buf, unsigned int bufLen)
 			m_resultPending = true;
 		}
 	}
+	return !m_resultPending;
 }
 
 void WindowsPipeReceiver::GetWaitObjects(WaitObjectContainer &container)
@@ -163,7 +165,8 @@ void WindowsPipeSender::Send(const byte* buf, unsigned int bufLen)
 {
 	DWORD written = 0;
 	HANDLE h = GetHandle();
-	if (WriteFile(h, buf, bufLen, &written, &m_overlapped))
+	// don't queue too much at once, or we might use up non-paged memory
+	if (WriteFile(h, buf, STDMIN(bufLen, 128U*1024U), &written, &m_overlapped))
 	{
 		m_resultPending = false;
 		m_lastResult = written;
diff --git a/winpipes.h b/winpipes.h
index 3260e196..29a3717a 100644
--- a/winpipes.h
+++ b/winpipes.h
@@ -7,7 +7,7 @@
 
 #include "network.h"
 #include "queue.h"
-#include <windows.h>
+#include <winsock2.h>
 
 NAMESPACE_BEGIN(CryptoPP)
 
@@ -64,7 +64,7 @@ public:
 	WindowsPipeReceiver();
 
 	bool MustWaitForResult() {return true;}
-	void Receive(byte* buf, unsigned int bufLen);
+	bool Receive(byte* buf, unsigned int bufLen);
 	unsigned int GetReceiveResult();
 	bool EofReceived() const {return m_eofReceived;}