diff --git a/README.build b/README.build
index c277b23f..d0c22e9c 100644
--- a/README.build
+++ b/README.build
@@ -50,11 +50,6 @@ Configure script options
 
 There are several other possible configure options
 
---enable-nofork		disables the fork into the background API
-			and removes all references to fork() and
-			pr_ctl() from the sources.  Use it if your
-			platform doesn't support forking.
-
 --enable-libcrypto	by default libwebsockets uses its own
 			built-in md5 and sha-1 implementation for
 			simplicity.  However the libcrypto ones
diff --git a/README.coding b/README.coding
index 05bc09fc..dfe6a414 100644
--- a/README.coding
+++ b/README.coding
@@ -25,24 +25,46 @@ similar to change the avaiable number of file descriptors, and when restarted
 libwebsockets will adapt accordingly.
 
 
-Procedure for sending data from other threads or process contexts
------------------------------------------------------------------
+Libwebsockets is singlethreaded
+-------------------------------
 
-Libwebsockets is carefully designed to work with no blocking in a single thread.
-In some cases where you will add libwebsockets to something else that uses the
-same single thread approach, you can so a safe implementation by combining the
-poll() loops as described in "External Polling loop support" below.
+Directly performing websocket actions from other threads is not allowed.
+Aside from the internal data being inconsistent in forked() processes,
+the scope of a wsi (struct websocket) can end at any time during service
+with the socket closing and the wsi freed.
 
-In other cases, you find you have asynchronous events coming from other thread
-or process contexts and there's not much you can do about it.  If you just try
-to randomly send, or broadcast using libwebsockets_broadcast() from these other
-places things will blow up either quickly or when the events on the two threads
-interefere with each other.  It's not legal to do this.
+Websocket write activities should only take place in the
+"LWS_CALLBACK_SERVER_WRITEABLE" callback as described below.
 
-For those situations, you can use libwebsockets_broadcast_foreign().  This
-serializes the data you're sending using a private, per-protocol socket, so the
-service thread picks it up when it's ready, and it is serviced from the service
-thread context only.
+Only live connections appear in the user callbacks, so this removes any
+possibility of trying to used closed and freed wsis.
+
+If you need to service other socket or file descriptors as well as the
+websocket ones, you can combine them together with the websocket ones
+in one poll loop, see "External Polling Loop support" below, and
+still do it all in one thread / process context.
+
+
+Only send data when socket writeable
+------------------------------------
+
+You should only send data on a websocket connection from the user callback
+"LWS_CALLBACK_SERVER_WRITEABLE" (or "LWS_CALLBACK_CLIENT_WRITEABLE" for
+clients).
+
+If you want to send something, do not just send it but request a callback
+when the socket is writeable using
+
+ - libwebsocket_callback_on_writable(context, wsi) for a specific wsi, or
+ - libwebsocket_callback_on_writable_all_protocol(protocol) for all connections
+using that protocol to get a callback when next writeable.
+
+Usually you will get called back immediately next time around the service
+loop, but if your peer is slow or temporarily inactive the callback will be
+delayed accordingly.  Generating what to write and sending it should be done
+in the ...WRITEABLE callback.
+
+See the test server code for an example of how to do this.
 
 
 Fragmented messages
diff --git a/README.test-apps b/README.test-apps
index bcc295f4..57afeb75 100644
--- a/README.test-apps
+++ b/README.test-apps
@@ -58,17 +58,6 @@ test-server.c is all that is needed to use libwebsockets for
 serving both the script html over http and websockets.
 
 
-Forkless operation
-------------------
-
-If your target device does not offer fork(), you can use
-libwebsockets from your own main loop instead.  Use the
-configure option --nofork and simply call libwebsocket_service()
-from your own main loop as shown in the test app sources.
-
-
-
-
 Testing websocket client support
 --------------------------------
 
diff --git a/configure.ac b/configure.ac
index 5cc5554d..d6868ec1 100644
--- a/configure.ac
+++ b/configure.ac
@@ -31,20 +31,6 @@ AC_CHECK_LIB([ssl], [SSL_library_init])
 CFLAGS="$CFLAGS -DLWS_OPENSSL_SUPPORT"
 fi
 
-#
-#
-#
-AC_ARG_ENABLE(nofork,
- [  --enable-nofork  Disables fork-related options],
- [ nofork=yes
- ])
-
-if test "x$nofork" = "xyes" ; then
-CFLAGS="$CFLAGS -DLWS_NO_FORK"
-else
-AC_FUNC_FORK
-fi
-
 #
 #
 #
diff --git a/lib/client.c b/lib/client.c
index 7a01a4ab..cf56b03d 100644
--- a/lib/client.c
+++ b/lib/client.c
@@ -627,7 +627,7 @@ libwebsockets_generate_client_handshake(struct libwebsocket_context *context,
 	int ext_count = 0;
 #endif
 	unsigned char buf[LWS_SEND_BUFFER_PRE_PADDING + 1 +
-			 MAX_BROADCAST_PAYLOAD + LWS_SEND_BUFFER_POST_PADDING];
+			 MAX_USER_RX_BUFFER + LWS_SEND_BUFFER_POST_PADDING];
 	static const char magic_websocket_guid[] =
 					 "258EAFA5-E914-47DA-95CA-C5AB0DC85B11";
 
diff --git a/lib/libwebsockets.c b/lib/libwebsockets.c
index 728d5e96..e6c87fd1 100644
--- a/lib/libwebsockets.c
+++ b/lib/libwebsockets.c
@@ -507,6 +507,7 @@ lws_handle_POLLOUT_event(struct libwebsocket_context *context,
 				struct libwebsocket *wsi, struct pollfd *pollfd)
 {
 	int n;
+
 #ifndef LWS_NO_EXTENSIONS
 	struct lws_tokens eff_buf;
 	int ret;
@@ -686,11 +687,12 @@ libwebsocket_service_fd(struct libwebsocket_context *context,
 							  struct pollfd *pollfd)
 {
 	struct libwebsocket *wsi;
-	unsigned char buf[LWS_SEND_BUFFER_PRE_PADDING + 1 +
-			 MAX_BROADCAST_PAYLOAD + LWS_SEND_BUFFER_POST_PADDING];
 	int n;
 	int m;
 	struct timeval tv;
+	unsigned char buf[LWS_SEND_BUFFER_PRE_PADDING + 1 +
+			 MAX_USER_RX_BUFFER + LWS_SEND_BUFFER_POST_PADDING];
+
 #ifndef LWS_NO_EXTENSIONS
 	int more = 1;
 #endif
@@ -781,11 +783,7 @@ libwebsocket_service_fd(struct libwebsocket_context *context,
 #ifndef LWS_NO_SERVER
 	case LWS_CONNMODE_HTTP_SERVING:
 	case LWS_CONNMODE_SERVER_LISTENER:
-	case LWS_CONNMODE_BROADCAST_PROXY_LISTENER:
-	case LWS_CONNMODE_BROADCAST_PROXY:
 	case LWS_CONNMODE_SSL_ACK_PENDING:
-
-lwsl_debug("*\n");
 		return lws_server_socket_service(context, wsi, pollfd);
 #endif
 
@@ -1001,10 +999,7 @@ libwebsocket_context_user(struct libwebsocket_context *context)
  *
  *	1) Accept new connections to our context's server
  *
- *	2) Perform pending broadcast writes initiated from other forked
- *	   processes (effectively serializing asynchronous broadcasts)
- *
- *	3) Call the receive callback for incoming frame data received by
+ *	2) Call the receive callback for incoming frame data received by
  *	    server or client connections.
  *
  *	You need to call this service function periodically to all the above
@@ -1423,11 +1418,6 @@ libwebsocket_create_context(int port, const char *interf,
 	struct sockaddr_in serv_addr;
 	int opt = 1;
 	struct libwebsocket_context *context = NULL;
-#ifndef LWS_NO_FORK
-	unsigned int slen;
-	struct sockaddr_in cli_addr;
-	int fd;
-#endif
 	char *p;
 	struct libwebsocket *wsi;
 #ifndef LWS_NO_EXTENSIONS
@@ -1445,7 +1435,6 @@ libwebsocket_create_context(int port, const char *interf,
 	lwsl_info(" LWS_INITIAL_HDR_ALLOC: %u\n", LWS_INITIAL_HDR_ALLOC);
 	lwsl_info(" LWS_ADDITIONAL_HDR_ALLOC: %u\n", LWS_ADDITIONAL_HDR_ALLOC);
 	lwsl_info(" MAX_USER_RX_BUFFER: %u\n", MAX_USER_RX_BUFFER);
-	lwsl_info(" MAX_BROADCAST_PAYLOAD: %u\n", MAX_BROADCAST_PAYLOAD);
 	lwsl_info(" LWS_MAX_PROTOCOLS: %u\n", LWS_MAX_PROTOCOLS);
 #ifndef LWS_NO_EXTENSIONS
 	lwsl_info(" LWS_MAX_EXTENSIONS_ACTIVE: %u\n", LWS_MAX_EXTENSIONS_ACTIVE);
@@ -1868,7 +1857,7 @@ libwebsocket_create_context(int port, const char *interf,
 			lwsl_warn("setuid: %s\n", strerror(errno));
 #endif
 
-	/* set up our internal broadcast trigger sockets per-protocol */
+	/* initialize supported protocols */
 
 	for (context->count_protocols = 0;
 			protocols[context->count_protocols].callback;
@@ -1880,64 +1869,6 @@ libwebsocket_create_context(int port, const char *interf,
 		protocols[context->count_protocols].owning_server = context;
 		protocols[context->count_protocols].protocol_index =
 						       context->count_protocols;
-
-#ifndef LWS_NO_FORK
-		fd = socket(AF_INET, SOCK_STREAM, 0);
-		if (fd < 0) {
-			lwsl_err("ERROR opening socket\n");
-			return NULL;
-		}
-
-		/* allow us to restart even if old sockets in TIME_WAIT */
-		setsockopt(fd, SOL_SOCKET, SO_REUSEADDR, (const void *)&opt,
-								  sizeof(opt));
-
-		bzero((char *) &serv_addr, sizeof(serv_addr));
-		serv_addr.sin_family = AF_INET;
-		serv_addr.sin_addr.s_addr = inet_addr("127.0.0.1");
-		serv_addr.sin_port = 0; /* pick the port for us */
-
-		n = bind(fd, (struct sockaddr *) &serv_addr, sizeof(serv_addr));
-		if (n < 0) {
-			lwsl_err("ERROR on binding to port %d (%d %d)\n",
-								port, n, errno);
-			return NULL;
-		}
-
-		slen = sizeof cli_addr;
-		n = getsockname(fd, (struct sockaddr *)&cli_addr, &slen);
-		if (n < 0) {
-			lwsl_err("getsockname failed\n");
-			return NULL;
-		}
-		protocols[context->count_protocols].broadcast_socket_port =
-						       ntohs(cli_addr.sin_port);
-		listen(fd, 5);
-
-		lwsl_debug("  Protocol %s broadcast socket %d\n",
-				protocols[context->count_protocols].name,
-						      ntohs(cli_addr.sin_port));
-
-		/* dummy wsi per broadcast proxy socket */
-
-		wsi = (struct libwebsocket *)malloc(sizeof(struct libwebsocket));
-		if (wsi == NULL) {
-			lwsl_err("Out of mem\n");
-			close(fd);
-			return NULL;
-		}
-		memset(wsi, 0, sizeof (struct libwebsocket));
-		wsi->sock = fd;
-		wsi->mode = LWS_CONNMODE_BROADCAST_PROXY_LISTENER;
-#ifndef LWS_NO_EXTENSIONS
-		wsi->count_active_extensions = 0;
-#endif
-		/* note which protocol we are proxying */
-		wsi->protocol_index_for_broadcast_proxy =
-						       context->count_protocols;
-
-		insert_wsi_socket_into_fds(context, wsi);
-#endif
 	}
 
 #ifndef LWS_NO_EXTENSIONS
@@ -1963,76 +1894,14 @@ libwebsocket_create_context(int port, const char *interf,
 	return context;
 }
 
-
-#ifndef LWS_NO_FORK
-
-/**
- * libwebsockets_fork_service_loop() - Optional helper function forks off
- *				  a process for the websocket server loop.
- *				You don't have to use this but if not, you
- *				have to make sure you are calling
- *				libwebsocket_service periodically to service
- *				the websocket traffic
- * @context:	server context returned by creation function
- */
-
-int
-libwebsockets_fork_service_loop(struct libwebsocket_context *context)
-{
-	int n;
-
-	n = fork();
-	if (n < 0)
-		return n;
-
-	if (n) {
-
-		/* main process context */
-
-		return 0;
-	}
-
-#ifdef HAVE_SYS_PRCTL_H
-	/* we want a SIGHUP when our parent goes down */
-	signal(SIGHUP, SIG_DFL);
-	prctl(PR_SET_PDEATHSIG, SIGHUP);
-#endif
-
-	/* in this forked process, sit and service websocket connections */
-
-	while (1) {
-		if (libwebsocket_service(context, 1000))
-			break;
-//#ifndef HAVE_SYS_PRCTL_H
-/*
- * on systems without prctl() (i.e. anything but linux) we can notice that our
- * parent is dead if getppid() returns 1. FIXME apparently this is not true for
- * solaris, could remember ppid right after fork and wait for it to change.
- */
-
-		/* if our parent went down, don't linger around */
-		if (context->started_with_parent && kill(context->started_with_parent, 0) < 0)
-			kill(getpid(), SIGTERM);
-
-	        if (getppid() == 1)
-	            break;
-//#endif
-	}
-
-
-	return 1;
-}
-
-#endif
-
 /**
  * libwebsockets_get_protocol() - Returns a protocol pointer from a websocket
  *				  connection.
  * @wsi:	pointer to struct websocket you want to know the protocol of
  *
  *
- *	This is useful to get the protocol to broadcast back to from inside
- * the callback.
+ *	Some apis can act on all live connections of a given protocol,
+ *	this is how you can get a pointer to the active protocol if needed.
  */
 
 const struct libwebsocket_protocols *
@@ -2041,159 +1910,6 @@ libwebsockets_get_protocol(struct libwebsocket *wsi)
 	return wsi->protocol;
 }
 
-/**
- * libwebsockets_broadcast() - Sends a buffer to tx callback for all connections of given protocol from single thread
- *
- * @protocol:	pointer to the protocol you will broadcast to all members of
- * @buf:  buffer containing the data to be broadcase.  NOTE: this has to be
- *		allocated with LWS_SEND_BUFFER_PRE_PADDING valid bytes before
- *		the pointer and LWS_SEND_BUFFER_POST_PADDING afterwards in the
- *		case you are calling this function from callback context.
- * @len:	length of payload data in buf, starting from buf.
- *
- *	This function allows bulk sending of a packet to every connection using
- * the given protocol.  It does not send the data directly; instead it calls
- * the callback with a reason type of LWS_CALLBACK_BROADCAST.  If the callback
- * wants to actually send the data for that connection, the callback itself
- * should call libwebsocket_write().
- *
- * This version only works from the same thread / process context as the service
- * loop.  Use libwesockets_broadcast_foreign(...) to do the same job from a different
- * thread in a safe way.
- */
-
-
-int
-libwebsockets_broadcast(const struct libwebsocket_protocols *protocol,
-						 unsigned char *buf, size_t len)
-{
-	struct libwebsocket_context *context = protocol->owning_server;
-	int n;
-	struct libwebsocket *wsi;
-
-	if (!context)
-		return 1;
-
-	/*
-	 * We are either running unforked / flat, or we are being
-	 * called from poll thread context
-	 * eg, from a callback.  In that case don't use sockets for
-	 * broadcast IPC (since we can't open a socket connection to
-	 * a socket listening on our own thread) but directly do the
-	 * send action.
-	 *
-	 * Locking is not needed because we are by definition being
-	 * called in the poll thread context and are serialized.
-	 */
-
-	for (n = 0; n < context->fds_count; n++) {
-
-		wsi = context->lws_lookup[context->fds[n].fd];
-		if (!wsi)
-			continue;
-
-		if (wsi->mode != LWS_CONNMODE_WS_SERVING)
-			continue;
-
-		/*
-		 * never broadcast to non-established connections
-		 */
-		if (wsi->state != WSI_STATE_ESTABLISHED)
-			continue;
-
-		/* only broadcast to guys using
-		 * requested protocol
-		 */
-		if (wsi->protocol != protocol)
-			continue;
-
-		user_callback_handle_rxflow(wsi->protocol->callback,
-			 context, wsi,
-			 LWS_CALLBACK_BROADCAST,
-			 wsi->user_space,
-			 buf, len);
-	}
-
-	return 0;
-}
-
-
-#ifndef LWS_NO_FORK
-/**
- * libwebsockets_broadcast_foreign() - Sends a buffer to the callback for all active
- *				  connections of the given protocol.
- * @protocol:	pointer to the protocol you will broadcast to all members of
- * @buf:  buffer containing the data to be broadcase.  NOTE: this has to be
- *		allocated with LWS_SEND_BUFFER_PRE_PADDING valid bytes before
- *		the pointer and LWS_SEND_BUFFER_POST_PADDING afterwards in the
- *		case you are calling this function from callback context.
- * @len:	length of payload data in buf, starting from buf.
- *
- *	This function allows bulk sending of a packet to every connection using
- * the given protocol.  It does not send the data directly; instead it calls
- * the callback with a reason type of LWS_CALLBACK_BROADCAST.  If the callback
- * wants to actually send the data for that connection, the callback itself
- * should call libwebsocket_write().
- *
- * This ..._foreign() version is designed to be randomly called from other thread or
- * process contexts than the main libwebsocket service one.  A private socket is used
- * to serialize accesses here with the main service loop. 
- */
-
-int
-libwebsockets_broadcast_foreign(struct libwebsocket_protocols *protocol,
-						 unsigned char *buf, size_t len)
-{
-	struct libwebsocket_context *context = protocol->owning_server;
-	int n;
-	int fd;
-	struct sockaddr_in cli_addr;
-
-	if (!context)
-		return 1;
-
-	/*
-	 * We're being called from a different process context than the server
-	 * loop.  Instead of broadcasting directly, we send our
-	 * payload on a socket to do the IPC; the server process will serialize
-	 * the broadcast action in its main poll() loop.
-	 *
-	 * There's one broadcast socket listening for each protocol supported
-	 * set up when the websocket server initializes
-	 */
-
-
-	/*
-	 * autoconnect to this protocol's broadcast proxy socket for this
-	 * thread if needed
-	 */
-
-	if (protocol->broadcast_socket_user_fd <= 0) {
-		fd = socket(AF_INET, SOCK_STREAM, 0);
-		if (fd < 0) {
-			lwsl_err("Unable to create socket\n");
-			return -1;
-		}
-		cli_addr.sin_family = AF_INET;
-		cli_addr.sin_port = htons(protocol->broadcast_socket_port);
-		cli_addr.sin_addr.s_addr = inet_addr("127.0.0.1");
-		n = connect(fd, (struct sockaddr *)&cli_addr, sizeof cli_addr);
-		if (n < 0) {
-			lwsl_err("Unable to connect to "
-					"broadcast socket %d, %s\n",
-					n, strerror(errno));
-			return -1;
-		}
-
-		protocol->broadcast_socket_user_fd = fd;
-	}
-
-	n = send(protocol->broadcast_socket_user_fd, buf, len, MSG_NOSIGNAL);
-
-	return n;
-}
-#endif
-
 int
 libwebsocket_is_final_fragment(struct libwebsocket *wsi)
 {
@@ -2224,39 +1940,6 @@ libwebsocket_ensure_user_space(struct libwebsocket *wsi)
 	return wsi->user_space;
 }
 
-/**
- * lws_confirm_legit_wsi: returns nonzero if the wsi looks bad
- *
- * @wsi: struct libwebsocket to assess
- *
- * Performs consistecy checks on what the wsi claims and what the
- * polling arrays hold.  This'll catch a closed wsi still in use.
- * Don't try to use on the listen (nonconnection) wsi as it will
- * fail it.  Otherwise 0 return == wsi seems consistent.
- */
-
-int lws_confirm_legit_wsi(struct libwebsocket *wsi)
-{
-	struct libwebsocket_context *context;
-
-	if (!(wsi && wsi->protocol && wsi->protocol->owning_server))
-		return 1;
-
-	context = wsi->protocol->owning_server;
-
-	if (!context)
-		return 2;
-
-	if (!wsi->position_in_fds_table)
-		return 3; /* position in fds table looks bad */
-	if (context->fds[wsi->position_in_fds_table].fd != wsi->sock)
-		return 4; /* pollfd entry does not wait on our socket descriptor */
-	if (context->lws_lookup[wsi->sock] != wsi)
-		return 5; /* lookup table does not agree with wsi */
-
-	return 0;
-}
-
 static void lwsl_emit_stderr(int level, const char *line)
 {
 	char buf[300];
diff --git a/lib/libwebsockets.h b/lib/libwebsockets.h
index 4d3342c4..0b08fcb2 100644
--- a/lib/libwebsockets.h
+++ b/lib/libwebsockets.h
@@ -725,9 +725,6 @@ libwebsocket_create_context(int port, const char * interf,
 LWS_EXTERN void
 libwebsocket_context_destroy(struct libwebsocket_context *context);
 
-LWS_EXTERN int
-libwebsockets_fork_service_loop(struct libwebsocket_context *context);
-
 LWS_EXTERN int
 libwebsocket_service(struct libwebsocket_context *context, int timeout_ms);
 
diff --git a/lib/output.c b/lib/output.c
index 69bf4593..5bf04043 100644
--- a/lib/output.c
+++ b/lib/output.c
@@ -286,10 +286,6 @@ int libwebsocket_write(struct libwebsocket *wsi, unsigned char *buf,
 	int m;
 #endif
 
-	if (lws_confirm_legit_wsi(wsi)) {
-		lwsl_err("libwebsocket_write on illegitimate wsi\n");
-		return -1;
-	}
 	if (len == 0 && protocol != LWS_WRITE_CLOSE) {
 		lwsl_warn("zero length libwebsocket_write attempt\n");
 		return 0;
diff --git a/lib/private-libwebsockets.h b/lib/private-libwebsockets.h
index 7e42202e..0e013b3c 100644
--- a/lib/private-libwebsockets.h
+++ b/lib/private-libwebsockets.h
@@ -107,9 +107,6 @@
 #ifndef MAX_USER_RX_BUFFER
 #define MAX_USER_RX_BUFFER 4096
 #endif
-#ifndef MAX_BROADCAST_PAYLOAD
-#define MAX_BROADCAST_PAYLOAD 4096
-#endif
 #ifndef LWS_MAX_PROTOCOLS
 #define LWS_MAX_PROTOCOLS 10
 #endif
@@ -217,8 +214,6 @@ enum connection_mode {
 
 	/* special internal types */
 	LWS_CONNMODE_SERVER_LISTENER,
-	LWS_CONNMODE_BROADCAST_PROXY_LISTENER,
-	LWS_CONNMODE_BROADCAST_PROXY
 };
 
 struct libwebsocket_protocols;
@@ -356,9 +351,6 @@ struct libwebsocket {
 	struct lws_tokens utf8_token[WSI_TOKEN_COUNT];
 	
 	enum libwebsocket_write_protocol rx_frame_type;
-#ifndef LWS_NO_FORK
-	int protocol_index_for_broadcast_proxy;
-#endif
 
 #ifndef LWS_NO_CLIENT
 	char *c_path;
@@ -449,9 +441,6 @@ lws_issue_raw_ext_access(struct libwebsocket *wsi,
 extern int
 _libwebsocket_rx_flow_control(struct libwebsocket *wsi);
 
-extern int
-lws_confirm_legit_wsi(struct libwebsocket *wsi);
-
 extern int
 user_callback_handle_rxflow(callback_function, struct libwebsocket_context * context,
 			struct libwebsocket *wsi,
diff --git a/lib/server.c b/lib/server.c
index 9aa1fe9f..0a7c0ddb 100644
--- a/lib/server.c
+++ b/lib/server.c
@@ -134,7 +134,7 @@ int lws_server_socket_service(struct libwebsocket_context *context,
 			struct libwebsocket *wsi, struct pollfd *pollfd)
 {
 	unsigned char buf[LWS_SEND_BUFFER_PRE_PADDING + 1 +
-			 MAX_BROADCAST_PAYLOAD + LWS_SEND_BUFFER_POST_PADDING];
+			 MAX_USER_RX_BUFFER + LWS_SEND_BUFFER_POST_PADDING];
 	struct libwebsocket *new_wsi;
 	int accept_fd;
 	unsigned int clilen;
@@ -368,130 +368,6 @@ int lws_server_socket_service(struct libwebsocket_context *context,
 		break;
 #endif
 
-
-#ifndef LWS_NO_FORK
-	case LWS_CONNMODE_BROADCAST_PROXY_LISTENER:
-
-		/* as we are listening, POLLIN means accept() is needed */
-
-		if (!(pollfd->revents & POLLIN))
-			break;
-
-		/* listen socket got an unencrypted connection... */
-
-		clilen = sizeof(cli_addr);
-		accept_fd  = accept(pollfd->fd, (struct sockaddr *)&cli_addr,
-								       &clilen);
-		if (accept_fd < 0) {
-			lwsl_warn("ERROR on accept %d\n", accept_fd);
-			return 0;
-		}
-
-		/* create a dummy wsi for the connection and add it */
-
-		new_wsi = (struct libwebsocket *)malloc(sizeof(struct libwebsocket));
-		if (new_wsi == NULL) {
-			lwsl_err("Out of mem\n");
-			goto bail_prox_listener;
-		}
-		memset(new_wsi, 0, sizeof (struct libwebsocket));
-		new_wsi->sock = accept_fd;
-		new_wsi->mode = LWS_CONNMODE_BROADCAST_PROXY;
-		new_wsi->state = WSI_STATE_ESTABLISHED;
-#ifndef LWS_NO_EXTENSIONS
-		new_wsi->count_active_extensions = 0;
-#endif
-		/* note which protocol we are proxying */
-		new_wsi->protocol_index_for_broadcast_proxy =
-					wsi->protocol_index_for_broadcast_proxy;
-
-		insert_wsi_socket_into_fds(context, new_wsi);
-		break;
-
-bail_prox_listener:
-		compatible_close(accept_fd);
-		break;
-
-	case LWS_CONNMODE_BROADCAST_PROXY:
-
-		/* handle session socket closed */
-
-		if (pollfd->revents & (POLLERR | POLLHUP)) {
-
-			lwsl_debug("Session Socket %p (fd=%d) dead\n",
-				(void *)wsi, pollfd->fd);
-
-			libwebsocket_close_and_free_session(context, wsi,
-						       LWS_CLOSE_STATUS_NORMAL);
-			return 0;
-		}
-
-		/*
-		 * either extension code with stuff to spill, or the user code,
-		 * requested a callback when it was OK to write
-		 */
-
-		if (pollfd->revents & POLLOUT)
-			if (lws_handle_POLLOUT_event(context, wsi,
-								 pollfd) < 0) {
-				libwebsocket_close_and_free_session(
-					context, wsi, LWS_CLOSE_STATUS_NORMAL);
-				return 0;
-			}
-
-		/* any incoming data ready? */
-
-		if (!(pollfd->revents & POLLIN))
-			break;
-
-		/* get the issued broadcast payload from the socket */
-
-		len = read(pollfd->fd, buf + LWS_SEND_BUFFER_PRE_PADDING,
-							 MAX_BROADCAST_PAYLOAD);
-		if (len < 0) {
-			lwsl_err("Error reading broadcast payload\n");
-			break;
-		}
-
-		/* broadcast it to all guys with this protocol index */
-
-		for (n = 0; n < context->fds_count; n++) {
-
-			new_wsi = context->lws_lookup[context->fds[n].fd];
-			if (new_wsi == NULL)
-				continue;
-
-			/* only to clients we are serving to */
-
-			if (new_wsi->mode != LWS_CONNMODE_WS_SERVING)
-				continue;
-
-			/*
-			 * never broadcast to non-established
-			 * connection
-			 */
-
-			if (new_wsi->state != WSI_STATE_ESTABLISHED)
-				continue;
-
-			/*
-			 * only broadcast to connections using
-			 * the requested protocol
-			 */
-
-			if (new_wsi->protocol->protocol_index !=
-				wsi->protocol_index_for_broadcast_proxy)
-				continue;
-
-			/* broadcast it to this connection */
-
-			user_callback_handle_rxflow(new_wsi->protocol->callback, context, new_wsi,
-				LWS_CALLBACK_BROADCAST,
-				new_wsi->user_space,
-				buf + LWS_SEND_BUFFER_PRE_PADDING, len);
-		}
-		break;
-#endif
 	default:
 		break;
 	}
diff --git a/libwebsockets-api-doc.html b/libwebsockets-api-doc.html
index 156afe79..402c36db 100644
--- a/libwebsockets-api-doc.html
+++ b/libwebsockets-api-doc.html
@@ -204,10 +204,7 @@ types of connection the same.
 <p>
 1) Accept new connections to our context's server
 <p>
-2) Perform pending broadcast writes initiated from other forked
-processes (effectively serializing asynchronous broadcasts)
-<p>
-3) Call the receive callback for incoming frame data received by
+2) Call the receive callback for incoming frame data received by
 server or client connections.
 <p>
 You need to call this service function periodically to all the above
@@ -400,16 +397,6 @@ images or whatever over http and dynamic data over websockets all in
 one place; they're all handled in the user callback.
 </blockquote>
 <hr>
-<h2>libwebsockets_fork_service_loop - Optional helper function forks off a process for the websocket server loop. You don't have to use this but if not, you have to make sure you are calling libwebsocket_service periodically to service the websocket traffic</h2>
-<i>int</i>
-<b>libwebsockets_fork_service_loop</b>
-(<i>struct libwebsocket_context *</i> <b>context</b>)
-<h3>Arguments</h3>
-<dl>
-<dt><b>context</b>
-<dd>server context returned by creation function
-</dl>
-<hr>
 <h2>libwebsockets_get_protocol - Returns a protocol pointer from a websocket connection.</h2>
 <i>const struct libwebsocket_protocols *</i>
 <b>libwebsockets_get_protocol</b>
@@ -422,87 +409,8 @@ one place; they're all handled in the user callback.
 <h3>Description</h3>
 <blockquote>
 <p>
-This is useful to get the protocol to broadcast back to from inside
-the callback.
-</blockquote>
-<hr>
-<h2>libwebsockets_broadcast - Sends a buffer to tx callback for all connections of given protocol from single thread</h2>
-<i>int</i>
-<b>libwebsockets_broadcast</b>
-(<i>const struct libwebsocket_protocols *</i> <b>protocol</b>,
-<i>unsigned char *</i> <b>buf</b>,
-<i>size_t</i> <b>len</b>)
-<h3>Arguments</h3>
-<dl>
-<dt><b>protocol</b>
-<dd>pointer to the protocol you will broadcast to all members of
-<dt><b>buf</b>
-<dd>buffer containing the data to be broadcase.  NOTE: this has to be
-allocated with LWS_SEND_BUFFER_PRE_PADDING valid bytes before
-the pointer and LWS_SEND_BUFFER_POST_PADDING afterwards in the
-case you are calling this function from callback context.
-<dt><b>len</b>
-<dd>length of payload data in buf, starting from buf.
-</dl>
-<h3>Description</h3>
-<blockquote>
-This function allows bulk sending of a packet to every connection using
-the given protocol.  It does not send the data directly; instead it calls
-the callback with a reason type of LWS_CALLBACK_BROADCAST.  If the callback
-wants to actually send the data for that connection, the callback itself
-should call <b>libwebsocket_write</b>.
-<p>
-This version only works from the same thread / process context as the service
-loop.  Use libwesockets_broadcast_foreign(...) to do the same job from a different
-thread in a safe way.
-</blockquote>
-<hr>
-<h2>libwebsockets_broadcast_foreign - Sends a buffer to the callback for all active connections of the given protocol.</h2>
-<i>int</i>
-<b>libwebsockets_broadcast_foreign</b>
-(<i>struct libwebsocket_protocols *</i> <b>protocol</b>,
-<i>unsigned char *</i> <b>buf</b>,
-<i>size_t</i> <b>len</b>)
-<h3>Arguments</h3>
-<dl>
-<dt><b>protocol</b>
-<dd>pointer to the protocol you will broadcast to all members of
-<dt><b>buf</b>
-<dd>buffer containing the data to be broadcase.  NOTE: this has to be
-allocated with LWS_SEND_BUFFER_PRE_PADDING valid bytes before
-the pointer and LWS_SEND_BUFFER_POST_PADDING afterwards in the
-case you are calling this function from callback context.
-<dt><b>len</b>
-<dd>length of payload data in buf, starting from buf.
-</dl>
-<h3>Description</h3>
-<blockquote>
-This function allows bulk sending of a packet to every connection using
-the given protocol.  It does not send the data directly; instead it calls
-the callback with a reason type of LWS_CALLBACK_BROADCAST.  If the callback
-wants to actually send the data for that connection, the callback itself
-should call <b>libwebsocket_write</b>.
-<p>
-This ...<b>_foreign</b> version is designed to be randomly called from other thread or
-process contexts than the main libwebsocket service one.  A private socket is used
-to serialize accesses here with the main service loop. 
-</blockquote>
-<hr>
-<h2>lws_confirm_legit_wsi - </h2>
-<i>int</i>
-<b>lws_confirm_legit_wsi</b>
-(<i>struct libwebsocket *</i> <b>wsi</b>)
-<h3>Arguments</h3>
-<dl>
-<dt><b>wsi</b>
-<dd>struct libwebsocket to assess
-</dl>
-<h3>Description</h3>
-<blockquote>
-Performs consistecy checks on what the wsi claims and what the
-polling arrays hold.  This'll catch a closed wsi still in use.
-Don't try to use on the listen (nonconnection) wsi as it will
-fail it.  Otherwise 0 return == wsi seems consistent.
+Some apis can act on all live connections of a given protocol,
+this is how you can get a pointer to the active protocol if needed.
 </blockquote>
 <hr>
 <h2>lws_set_log_level - Set the logging bitfield</h2>
diff --git a/test-server/test-server.c b/test-server/test-server.c
index 2fa0eae9..505c6eec 100644
--- a/test-server/test-server.c
+++ b/test-server/test-server.c
@@ -31,20 +31,12 @@
 #include "../lib/libwebsockets.h"
 
 static int close_testing;
-
-#ifdef EXTERNAL_POLL
-#ifndef LWS_NO_FORK
-#define LWS_NO_FORK
-#endif
-
 int max_poll_elements;
 
 struct pollfd *pollfds;
 int *fd_lookup;
 int count_pollfds;
 
-#endif /* EXTERNAL_POLL */
-
 
 /*
  * This demo server shows how to use libwebsockets for one or more
@@ -277,13 +269,7 @@ callback_dumb_increment(struct libwebsocket_context *context,
 		pss->number = 0;
 		break;
 
-	/*
-	 * in this protocol, we just use the broadcast action as the chance to
-	 * send our own connection-specific data and ignore the broadcast info
-	 * that is available in the 'in' parameter
-	 */
-
-	case LWS_CALLBACK_BROADCAST:
+	case LWS_CALLBACK_SERVER_WRITEABLE:
 		n = sprintf((char *)p, "%d", pss->number++);
 		n = libwebsocket_write(wsi, p, n, LWS_WRITE_TEXT);
 		if (n < 0) {
@@ -396,12 +382,6 @@ callback_lws_mirror(struct libwebsocket_context *context,
 		}
 		break;
 
-	case LWS_CALLBACK_BROADCAST:
-		n = libwebsocket_write(wsi, in, len, LWS_WRITE_TEXT);
-		if (n < 0)
-			lwsl_err("mirror write failed\n");
-		break;
-
 	case LWS_CALLBACK_RECEIVE:
 
 		if (((ringbuffer_head - pss->ringbuffer_tail) &
@@ -504,8 +484,6 @@ int main(int argc, char **argv)
 			    LOCAL_RESOURCE_PATH"/libwebsockets-test-server.pem";
 	const char *key_path =
 			LOCAL_RESOURCE_PATH"/libwebsockets-test-server.key.pem";
-	unsigned char buf[LWS_SEND_BUFFER_PRE_PADDING + 1024 +
-						  LWS_SEND_BUFFER_POST_PADDING];
 	int port = 7681;
 	int use_ssl = 0;
 	struct libwebsocket_context *context;
@@ -513,9 +491,8 @@ int main(int argc, char **argv)
 	char interface_name[128] = "";
 	const char *interface = NULL;
 	int syslog_options = LOG_PID | LOG_PERROR;
-#ifdef LWS_NO_FORK
 	unsigned int oldus = 0;
-#endif
+
 	int debug_level = 7;
 #ifndef LWS_NO_DAEMONIZE
 	int daemonize = 0;
@@ -605,16 +582,6 @@ int main(int argc, char **argv)
 		return -1;
 	}
 
-	buf[LWS_SEND_BUFFER_PRE_PADDING] = 'x';
-
-#ifdef LWS_NO_FORK
-
-	/*
-	 * This example shows how to work with no forked service loop
-	 */
-
-	lwsl_info(" Using no-fork service loop\n");
-
 	n = 0;
 	while (n >= 0) {
 		struct timeval tv;
@@ -622,22 +589,13 @@ int main(int argc, char **argv)
 		gettimeofday(&tv, NULL);
 
 		/*
-		 * This broadcasts to all dumb-increment-protocol connections
-		 * at 20Hz.
-		 *
-		 * We're just sending a character 'x', in these examples the
-		 * callbacks send their own per-connection content.
-		 *
-		 * You have to send something with nonzero length to get the
-		 * callback actions delivered.
-		 *
-		 * We take care of pre-and-post padding allocation.
+		 * This provokes the LWS_CALLBACK_SERVER_WRITEABLE for every
+		 * live websocket connection using the DUMB_INCREMENT protocol,
+		 * as soon as it can take more packets (usually immediately)
 		 */
 
 		if (((unsigned int)tv.tv_usec - oldus) > 50000) {
-			libwebsockets_broadcast(
-					&protocols[PROTOCOL_DUMB_INCREMENT],
-					&buf[LWS_SEND_BUFFER_PRE_PADDING], 1);
+			libwebsocket_callback_on_writable_all_protocol(&protocols[PROTOCOL_DUMB_INCREMENT]);
 			oldus = tv.tv_usec;
 		}
 
@@ -679,49 +637,6 @@ int main(int argc, char **argv)
 #endif
 	}
 
-#else /* !LWS_NO_FORK */
-
-	/*
-	 * This example shows how to work with the forked websocket service loop
-	 */
-
-	lwsl_info(" Using forked service loop\n");
-
-	/*
-	 * This forks the websocket service action into a subprocess so we
-	 * don't have to take care about it.
-	 */
-
-	n = libwebsockets_fork_service_loop(context);
-	if (n < 0) {
-		lwsl_err("Unable to fork service loop %d\n", n);
-		return 1;
-	}
-	if (n)
-		exit(0);
-
-	while (1) {
-
-		usleep(50000);
-
-		/*
-		 * This broadcasts to all dumb-increment-protocol connections
-		 * at 20Hz.
-		 *
-		 * We're just sending a character 'x', in these examples the
-		 * callbacks send their own per-connection content.
-		 *
-		 * You have to send something with nonzero length to get the
-		 * callback actions delivered.
-		 *
-		 * We take care of pre-and-post padding allocation.
-		 */
-
-		libwebsockets_broadcast_foreign(&protocols[PROTOCOL_DUMB_INCREMENT],
-					&buf[LWS_SEND_BUFFER_PRE_PADDING], 1);
-	}
-
-#endif
 #ifdef EXTERNAL_POLL
 done:
 #endif