diff --git a/mongoose.c b/mongoose.c
index 644f9220..21532c17 100644
--- a/mongoose.c
+++ b/mongoose.c
@@ -16,7 +16,930 @@
 // license, as set out in <http://cesanta.com/>.
 
 // net_skeleton start
-#include "net_skeleton.h"
+
+// Copyright (c) 2014 Cesanta Software Limited
+// All rights reserved
+//
+// This library is dual-licensed: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License version 2 as
+// published by the Free Software Foundation. For the terms of this
+// license, see <http://www.gnu.org/licenses/>.
+//
+// You are free to use this library under the terms of the GNU General
+// Public License, but WITHOUT ANY WARRANTY; without even the implied
+// warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+// See the GNU General Public License for more details.
+//
+// Alternatively, you can license this library under a commercial
+// license, as set out in <http://cesanta.com/>.
+
+#ifndef NS_SKELETON_HEADER_INCLUDED
+#define NS_SKELETON_HEADER_INCLUDED
+
+#define NS_SKELETON_VERSION "1.0"
+
+#undef UNICODE                  // Use ANSI WinAPI functions
+#undef _UNICODE                 // Use multibyte encoding on Windows
+#define _MBCS                   // Use multibyte encoding on Windows
+#define _INTEGRAL_MAX_BITS 64   // Enable _stati64() on Windows
+#define _CRT_SECURE_NO_WARNINGS // Disable deprecation warning in VS2005+
+#undef WIN32_LEAN_AND_MEAN      // Let windows.h always include winsock2.h
+#define _XOPEN_SOURCE 600       // For flockfile() on Linux
+#define __STDC_FORMAT_MACROS    // <inttypes.h> wants this for C++
+#define __STDC_LIMIT_MACROS     // C++ wants that for INT64_MAX
+#define _LARGEFILE_SOURCE       // Enable fseeko() and ftello() functions
+#define _FILE_OFFSET_BITS 64    // Enable 64-bit file offsets
+
+#ifdef _MSC_VER
+#pragma warning (disable : 4127)  // FD_SET() emits warning, disable it
+#pragma warning (disable : 4204)  // missing c99 support
+#endif
+
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <assert.h>
+#include <errno.h>
+#include <fcntl.h>
+#include <stdarg.h>
+#include <stddef.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <time.h>
+
+#ifdef _WIN32
+#pragma comment(lib, "ws2_32.lib")    // Linking with winsock library
+#include <windows.h>
+#include <process.h>
+#ifndef EINPROGRESS
+#define EINPROGRESS WSAEINPROGRESS
+#endif
+#ifndef EWOULDBLOCK
+#define EWOULDBLOCK WSAEWOULDBLOCK
+#endif
+#ifndef __func__
+#define STRX(x) #x
+#define STR(x) STRX(x)
+#define __func__ __FILE__ ":" STR(__LINE__)
+#endif
+#ifndef va_copy
+#define va_copy(x,y) x = y
+#endif // MINGW #defines va_copy
+#define snprintf _snprintf
+#define vsnprintf _vsnprintf
+#define to64(x) _atoi64(x)
+typedef int socklen_t;
+typedef unsigned char uint8_t;
+typedef unsigned int uint32_t;
+typedef unsigned short uint16_t;
+typedef unsigned __int64 uint64_t;
+typedef __int64   int64_t;
+typedef SOCKET sock_t;
+#else
+#include <errno.h>
+#include <fcntl.h>
+#include <netdb.h>
+#include <pthread.h>
+#include <stdarg.h>
+#include <unistd.h>
+#include <arpa/inet.h>  // For inet_pton() when NS_ENABLE_IPV6 is defined
+#include <netinet/in.h>
+#include <sys/socket.h>
+#include <sys/select.h>
+#define closesocket(x) close(x)
+#define __cdecl
+#define INVALID_SOCKET (-1)
+#define to64(x) strtoll(x, NULL, 10)
+typedef int sock_t;
+#endif
+
+#ifdef NS_ENABLE_DEBUG
+#define DBG(x) do { printf("%-20s ", __func__); printf x; putchar('\n'); \
+  fflush(stdout); } while(0)
+#else
+#define DBG(x)
+#endif
+
+#define ARRAY_SIZE(array) (sizeof(array) / sizeof(array[0]))
+
+#ifdef NS_ENABLE_SSL
+#ifdef __APPLE__
+#pragma GCC diagnostic ignored "-Wdeprecated-declarations"
+#endif
+#include <openssl/ssl.h>
+#else
+typedef void *SSL;
+typedef void *SSL_CTX;
+#endif
+
+#ifdef __cplusplus
+extern "C" {
+#endif // __cplusplus
+
+union socket_address {
+  struct sockaddr sa;
+  struct sockaddr_in sin;
+#ifdef NS_ENABLE_IPV6
+  struct sockaddr_in6 sin6;
+#endif
+};
+
+struct iobuf {
+  char *buf;
+  int len;
+  int size;
+};
+
+void iobuf_init(struct iobuf *, int initial_size);
+void iobuf_free(struct iobuf *);
+int iobuf_append(struct iobuf *, const void *data, int data_size);
+void iobuf_remove(struct iobuf *, int data_size);
+
+struct ns_connection;
+enum ns_event { NS_POLL, NS_ACCEPT, NS_CONNECT, NS_RECV, NS_SEND, NS_CLOSE };
+typedef void (*ns_callback_t)(struct ns_connection *, enum ns_event, void *);
+
+struct ns_server {
+  void *server_data;
+  union socket_address listening_sa;
+  sock_t listening_sock;
+  struct ns_connection *active_connections;
+  ns_callback_t callback;
+  SSL_CTX *ssl_ctx;
+  SSL_CTX *client_ssl_ctx;
+};
+
+struct ns_connection {
+  struct ns_connection *prev, *next;
+  struct ns_server *server;
+  void *connection_data;
+  time_t last_io_time;
+  sock_t sock;
+  struct iobuf recv_iobuf;
+  struct iobuf send_iobuf;
+  SSL *ssl;
+  unsigned int flags;
+#define NSF_FINISHED_SENDING_DATA   (1 << 0)
+#define NSF_BUFFER_BUT_DONT_SEND    (1 << 1)
+#define NSF_SSL_HANDSHAKE_DONE      (1 << 2)
+#define NSF_CONNECTING              (1 << 3)
+#define NSF_CLOSE_IMMEDIATELY       (1 << 4)
+#define NSF_ACCEPTED                (1 << 5)
+#define NSF_USER_1                  (1 << 6)
+#define NSF_USER_2                  (1 << 7)
+#define NSF_USER_3                  (1 << 8)
+#define NSF_USER_4                  (1 << 9)
+};
+
+void ns_server_init(struct ns_server *, void *server_data, ns_callback_t);
+void ns_server_free(struct ns_server *);
+int ns_server_poll(struct ns_server *, int milli);
+void ns_server_wakeup(struct ns_server *, void *conn_param);
+void ns_iterate(struct ns_server *, ns_callback_t cb, void *param);
+struct ns_connection *ns_add_sock(struct ns_server *, sock_t sock, void *p);
+
+int ns_bind(struct ns_server *, const char *addr);
+int ns_set_ssl_cert(struct ns_server *, const char *ssl_cert);
+struct ns_connection *ns_connect(struct ns_server *, const char *host,
+                                 int port, int ssl, void *connection_param);
+
+int ns_send(struct ns_connection *, const void *buf, int len);
+int ns_printf(struct ns_connection *, const char *fmt, ...);
+int ns_vprintf(struct ns_connection *, const char *fmt, va_list ap);
+
+// Utility functions
+void *ns_start_thread(void *(*f)(void *), void *p);
+int ns_socketpair(sock_t [2]);
+void ns_set_close_on_exec(sock_t);
+
+#ifdef __cplusplus
+}
+#endif // __cplusplus
+
+#endif // NS_SKELETON_HEADER_INCLUDED
+// Copyright (c) 2014 Cesanta Software Limited
+// All rights reserved
+//
+// This library is dual-licensed: you can redistribute it and/or modify
+// it under the terms of the GNU General Public License version 2 as
+// published by the Free Software Foundation. For the terms of this
+// license, see <http://www.gnu.org/licenses/>.
+//
+// You are free to use this library under the terms of the GNU General
+// Public License, but WITHOUT ANY WARRANTY; without even the implied
+// warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
+// See the GNU General Public License for more details.
+//
+// Alternatively, you can license this library under a commercial
+// license, as set out in <http://cesanta.com/>.
+
+
+#ifndef NS_MALLOC
+#define NS_MALLOC malloc
+#endif
+
+#ifndef NS_REALLOC
+#define NS_REALLOC realloc
+#endif
+
+#ifndef NS_FREE
+#define NS_FREE free
+#endif
+
+#ifndef IOBUF_RESIZE_MULTIPLIER
+#define IOBUF_RESIZE_MULTIPLIER 2.0
+#endif
+
+void iobuf_init(struct iobuf *iobuf, int size) {
+  iobuf->len = iobuf->size = 0;
+  iobuf->buf = NULL;
+
+  if (size > 0 && (iobuf->buf = (char *) NS_MALLOC(size)) != NULL) {
+    iobuf->size = size;
+  }
+}
+
+void iobuf_free(struct iobuf *iobuf) {
+  if (iobuf != NULL) {
+    if (iobuf->buf != NULL) NS_FREE(iobuf->buf);
+    iobuf_init(iobuf, 0);
+  }
+}
+
+int iobuf_append(struct iobuf *io, const void *buf, int len) {
+  static const double mult = IOBUF_RESIZE_MULTIPLIER;
+  char *p = NULL;
+  int new_len = 0;
+
+  assert(io->len >= 0);
+  assert(io->len <= io->size);
+
+  if (len <= 0) {
+  } else if ((new_len = io->len + len) < io->size) {
+    memcpy(io->buf + io->len, buf, len);
+    io->len = new_len;
+  } else if ((p = (char *)
+              NS_REALLOC(io->buf, (int) (new_len * mult))) != NULL) {
+    io->buf = p;
+    memcpy(io->buf + io->len, buf, len);
+    io->len = new_len;
+    io->size = (int) (new_len * mult);
+  } else {
+    len = 0;
+  }
+
+  return len;
+}
+
+void iobuf_remove(struct iobuf *io, int n) {
+  if (n >= 0 && n <= io->len) {
+    memmove(io->buf, io->buf + n, io->len - n);
+    io->len -= n;
+  }
+}
+
+#ifndef NS_DISABLE_THREADS
+void *ns_start_thread(void *(*f)(void *), void *p) {
+#ifdef _WIN32
+  return (void *) _beginthread((void (__cdecl *)(void *)) f, 0, p);
+#else
+  pthread_t thread_id = (pthread_t) 0;
+  pthread_attr_t attr;
+
+  (void) pthread_attr_init(&attr);
+  (void) pthread_attr_setdetachstate(&attr, PTHREAD_CREATE_DETACHED);
+
+#if NS_STACK_SIZE > 1
+  (void) pthread_attr_setstacksize(&attr, NS_STACK_SIZE);
+#endif
+
+  pthread_create(&thread_id, &attr, f, p);
+  pthread_attr_destroy(&attr);
+
+  return (void *) thread_id;
+#endif
+}
+#endif  // NS_DISABLE_THREADS
+
+static void add_connection(struct ns_server *server, struct ns_connection *c) {
+  c->next = server->active_connections;
+  server->active_connections = c;
+  c->prev = NULL;
+  if (c->next != NULL) c->next->prev = c;
+}
+
+static void remove_connection(struct ns_connection *conn) {
+  if (conn->prev == NULL) conn->server->active_connections = conn->next;
+  if (conn->prev) conn->prev->next = conn->next;
+  if (conn->next) conn->next->prev = conn->prev;
+}
+
+// Print message to buffer. If buffer is large enough to hold the message,
+// return buffer. If buffer is to small, allocate large enough buffer on heap,
+// and return allocated buffer.
+static int alloc_vprintf(char **buf, size_t size, const char *fmt, va_list ap) {
+  va_list ap_copy;
+  int len;
+
+  va_copy(ap_copy, ap);
+  len = vsnprintf(*buf, size, fmt, ap_copy);
+  va_end(ap_copy);
+
+  if (len < 0) {
+    // eCos and Windows are not standard-compliant and return -1 when
+    // the buffer is too small. Keep allocating larger buffers until we
+    // succeed or out of memory.
+    *buf = NULL;
+    while (len < 0) {
+      if (*buf) free(*buf);
+      size *= 2;
+      if ((*buf = (char *) NS_MALLOC(size)) == NULL) break;
+      va_copy(ap_copy, ap);
+      len = vsnprintf(*buf, size, fmt, ap_copy);
+      va_end(ap_copy);
+    }
+  } else if (len > (int) size) {
+    // Standard-compliant code path. Allocate a buffer that is large enough.
+    if ((*buf = (char *) NS_MALLOC(len + 1)) == NULL) {
+      len = -1;
+    } else {
+      va_copy(ap_copy, ap);
+      len = vsnprintf(*buf, len + 1, fmt, ap_copy);
+      va_end(ap_copy);
+    }
+  }
+
+  return len;
+}
+
+int ns_vprintf(struct ns_connection *conn, const char *fmt, va_list ap) {
+  char mem[2000], *buf = mem;
+  int len;
+
+  if ((len = alloc_vprintf(&buf, sizeof(mem), fmt, ap)) > 0) {
+    iobuf_append(&conn->send_iobuf, buf, len);
+  }
+  if (buf != mem && buf != NULL) {
+    free(buf);
+  }
+
+  return len;
+}
+
+int ns_printf(struct ns_connection *conn, const char *fmt, ...) {
+  int len;
+  va_list ap;
+  va_start(ap, fmt);
+  len = ns_vprintf(conn, fmt, ap);
+  va_end(ap);
+  return len;
+}
+
+static void call_user(struct ns_connection *conn, enum ns_event ev, void *p) {
+  if (conn->server->callback) conn->server->callback(conn, ev, p);
+}
+
+static void close_conn(struct ns_connection *conn) {
+  DBG(("%p %d", conn, conn->flags));
+  call_user(conn, NS_CLOSE, NULL);
+  remove_connection(conn);
+  closesocket(conn->sock);
+  iobuf_free(&conn->recv_iobuf);
+  iobuf_free(&conn->send_iobuf);
+  NS_FREE(conn);
+}
+
+void ns_set_close_on_exec(sock_t sock) {
+#ifdef _WIN32
+  (void) SetHandleInformation((HANDLE) sock, HANDLE_FLAG_INHERIT, 0);
+#else
+  fcntl(sock, F_SETFD, FD_CLOEXEC);
+#endif
+}
+
+static void set_non_blocking_mode(sock_t sock) {
+#ifdef _WIN32
+  unsigned long on = 1;
+  ioctlsocket(sock, FIONBIO, &on);
+#else
+  int flags = fcntl(sock, F_GETFL, 0);
+  fcntl(sock, F_SETFL, flags | O_NONBLOCK);
+#endif
+}
+
+#ifndef NS_DISABLE_SOCKETPAIR
+int ns_socketpair(sock_t sp[2]) {
+  struct sockaddr_in sa;
+  sock_t sock;
+  socklen_t len = sizeof(sa);
+  int ret = 0;
+
+  sp[0] = sp[1] = INVALID_SOCKET;
+
+  (void) memset(&sa, 0, sizeof(sa));
+  sa.sin_family = AF_INET;
+  sa.sin_port = htons(0);
+  sa.sin_addr.s_addr = htonl(0x7f000001);
+
+  if ((sock = socket(AF_INET, SOCK_STREAM, 0)) != INVALID_SOCKET &&
+      !bind(sock, (struct sockaddr *) &sa, len) &&
+      !listen(sock, 1) &&
+      !getsockname(sock, (struct sockaddr *) &sa, &len) &&
+      (sp[0] = socket(AF_INET, SOCK_STREAM, 6)) != -1 &&
+      !connect(sp[0], (struct sockaddr *) &sa, len) &&
+      (sp[1] = accept(sock,(struct sockaddr *) &sa, &len)) != INVALID_SOCKET) {
+    ns_set_close_on_exec(sp[0]);
+    ns_set_close_on_exec(sp[1]);
+    ret = 1;
+  } else {
+    if (sp[0] != INVALID_SOCKET) closesocket(sp[0]);
+    if (sp[1] != INVALID_SOCKET) closesocket(sp[1]);
+    sp[0] = sp[1] = INVALID_SOCKET;
+  }
+  closesocket(sock);
+
+  return ret;
+}
+#endif  // NS_DISABLE_SOCKETPAIR
+
+// Valid listening port spec is: [ip_address:]port, e.g. "80", "127.0.0.1:3128"
+static int parse_port_string(const char *str, union socket_address *sa) {
+  unsigned int a, b, c, d, port;
+  int len = 0;
+#ifdef NS_ENABLE_IPV6
+  char buf[100];
+#endif
+
+  // MacOS needs that. If we do not zero it, subsequent bind() will fail.
+  // Also, all-zeroes in the socket address means binding to all addresses
+  // for both IPv4 and IPv6 (INADDR_ANY and IN6ADDR_ANY_INIT).
+  memset(sa, 0, sizeof(*sa));
+  sa->sin.sin_family = AF_INET;
+
+  if (sscanf(str, "%u.%u.%u.%u:%u%n", &a, &b, &c, &d, &port, &len) == 5) {
+    // Bind to a specific IPv4 address, e.g. 192.168.1.5:8080
+    sa->sin.sin_addr.s_addr = htonl((a << 24) | (b << 16) | (c << 8) | d);
+    sa->sin.sin_port = htons((uint16_t) port);
+#ifdef NS_ENABLE_IPV6
+  } else if (sscanf(str, "[%49[^]]]:%u%n", buf, &port, &len) == 2 &&
+             inet_pton(AF_INET6, buf, &sa->sin6.sin6_addr)) {
+    // IPv6 address, e.g. [3ffe:2a00:100:7031::1]:8080
+    sa->sin6.sin6_family = AF_INET6;
+    sa->sin6.sin6_port = htons((uint16_t) port);
+#endif
+  } else if (sscanf(str, "%u%n", &port, &len) == 1) {
+    // If only port is specified, bind to IPv4, INADDR_ANY
+    sa->sin.sin_port = htons((uint16_t) port);
+  } else {
+    port = 0;   // Parsing failure. Make port invalid.
+  }
+
+  return port <= 0xffff && str[len] == '\0';
+}
+
+// 'sa' must be an initialized address to bind to
+static sock_t open_listening_socket(union socket_address *sa) {
+  socklen_t len = sizeof(*sa);
+  sock_t on = 1, sock = INVALID_SOCKET;
+
+  if ((sock = socket(sa->sa.sa_family, SOCK_STREAM, 6)) != INVALID_SOCKET &&
+      !setsockopt(sock, SOL_SOCKET, SO_REUSEADDR, (void *) &on, sizeof(on)) &&
+      !bind(sock, &sa->sa, sa->sa.sa_family == AF_INET ?
+            sizeof(sa->sin) : sizeof(sa->sa)) &&
+      !listen(sock, SOMAXCONN)) {
+    set_non_blocking_mode(sock);
+    // In case port was set to 0, get the real port number
+    (void) getsockname(sock, &sa->sa, &len);
+  } else if (sock != INVALID_SOCKET) {
+    closesocket(sock);
+    sock = INVALID_SOCKET;
+  }
+
+  return sock;
+}
+
+
+int ns_set_ssl_cert(struct ns_server *server, const char *cert) {
+#ifdef NS_ENABLE_SSL
+  if (cert != NULL &&
+      (server->ssl_ctx = SSL_CTX_new(SSLv23_server_method())) == NULL) {
+    return -1;
+  } else if (SSL_CTX_use_certificate_file(server->ssl_ctx, cert, 1) == 0 ||
+             SSL_CTX_use_PrivateKey_file(server->ssl_ctx, cert, 1) == 0) {
+    return -2;
+  } else {
+    SSL_CTX_use_certificate_chain_file(server->ssl_ctx, cert);
+  }
+  return 0;
+#else
+  return server != NULL && cert == NULL ? 0 : -3;
+#endif
+}
+
+int ns_bind(struct ns_server *server, const char *str) {
+  parse_port_string(str, &server->listening_sa);
+  if (server->listening_sock != INVALID_SOCKET) {
+    closesocket(server->listening_sock);
+  }
+  server->listening_sock = open_listening_socket(&server->listening_sa);
+  return server->listening_sock == INVALID_SOCKET ? -1 :
+    (int) ntohs(server->listening_sa.sin.sin_port);
+}
+
+
+static struct ns_connection *accept_conn(struct ns_server *server) {
+  struct ns_connection *c = NULL;
+  union socket_address sa;
+  socklen_t len = sizeof(sa);
+  sock_t sock = INVALID_SOCKET;
+
+  // NOTE(lsm): on Windows, sock is always > FD_SETSIZE
+  if ((sock = accept(server->listening_sock, &sa.sa, &len)) == INVALID_SOCKET) {
+    closesocket(sock);
+  } else if ((c = (struct ns_connection *) NS_MALLOC(sizeof(*c))) == NULL ||
+             memset(c, 0, sizeof(*c)) == NULL) {
+    closesocket(sock);
+#ifdef NS_ENABLE_SSL
+  } else if (server->ssl_ctx != NULL &&
+             ((c->ssl = SSL_new(server->ssl_ctx)) == NULL ||
+              SSL_set_fd(c->ssl, sock) != 1)) {
+    DBG(("SSL error"));
+    closesocket(sock);
+    free(c);
+    c = NULL;
+#endif
+  } else {
+    ns_set_close_on_exec(sock);
+    set_non_blocking_mode(sock);
+    c->server = server;
+    c->sock = sock;
+    c->flags |= NSF_ACCEPTED;
+
+    add_connection(server, c);
+    call_user(c, NS_ACCEPT, &sa);
+    DBG(("%p %d %p %p", c, c->sock, c->ssl, server->ssl_ctx));
+  }
+
+  return c;
+}
+
+static int ns_is_error(int n) {
+  return n == 0 ||
+    (n < 0 && errno != EINTR && errno != EINPROGRESS &&
+     errno != EAGAIN && errno != EWOULDBLOCK
+#ifdef _WIN32
+     && WSAGetLastError() != WSAEINTR && WSAGetLastError() != WSAEWOULDBLOCK
+#endif
+    );
+}
+
+#ifdef NS_ENABLE_HEXDUMP
+static void hexdump(const struct ns_connection *conn, const void *buf,
+                    int len, const char *marker) {
+  const unsigned char *p = (const unsigned char *) buf;
+  char path[500], date[100], ascii[17];
+  FILE *fp;
+
+#if 0
+  if (!match_prefix(NS_ENABLE_HEXDUMP, strlen(NS_ENABLE_HEXDUMP),
+                    conn->remote_ip)) {
+    return;
+  }
+
+  snprintf(path, sizeof(path), "%s.%hu.txt",
+           conn->mg_conn.remote_ip, conn->mg_conn.remote_port);
+#endif
+  snprintf(path, sizeof(path), "%p.txt", conn);
+
+  if ((fp = fopen(path, "a")) != NULL) {
+    time_t cur_time = time(NULL);
+    int i, idx;
+
+    strftime(date, sizeof(date), "%d/%b/%Y %H:%M:%S", localtime(&cur_time));
+    fprintf(fp, "%s %s %d bytes\n", marker, date, len);
+
+    for (i = 0; i < len; i++) {
+      idx = i % 16;
+      if (idx == 0) {
+        if (i > 0) fprintf(fp, "  %s\n", ascii);
+        fprintf(fp, "%04x ", i);
+      }
+      fprintf(fp, " %02x", p[i]);
+      ascii[idx] = p[i] < 0x20 || p[i] > 0x7e ? '.' : p[i];
+      ascii[idx + 1] = '\0';
+    }
+
+    while (i++ % 16) fprintf(fp, "%s", "   ");
+    fprintf(fp, "  %s\n\n", ascii);
+
+    fclose(fp);
+  }
+}
+#endif
+
+static void read_from_socket(struct ns_connection *conn) {
+  char buf[2048];
+  int n = 0;
+
+  if (conn->flags & NSF_CONNECTING) {
+    int ok = 1, ret;
+    socklen_t len = sizeof(ok);
+
+    conn->flags &= ~NSF_CONNECTING;
+    ret = getsockopt(conn->sock, SOL_SOCKET, SO_ERROR, (char *) &ok, &len);
+#ifdef NS_ENABLE_SSL
+    if (ret == 0 && ok == 0 && conn->ssl != NULL) {
+      int res = SSL_connect(conn->ssl);
+      int ssl_err = SSL_get_error(conn->ssl, res);
+      DBG(("%p res %d %d", conn, res, ssl_err));
+      if (res == 1) {
+        conn->flags = NSF_SSL_HANDSHAKE_DONE;
+      } else if (res == 0 || ssl_err == 2 || ssl_err == 3) {
+        conn->flags |= NSF_CONNECTING;
+        return; // Call us again
+      } else {
+        ok = 1;
+      }
+    }
+#endif
+    DBG(("%p ok=%d", conn, ok));
+    if (ok != 0) {
+      conn->flags |= NSF_CLOSE_IMMEDIATELY;
+    }
+    call_user(conn, NS_CONNECT, &ok);
+    return;
+  }
+
+#ifdef NS_ENABLE_SSL
+  if (conn->ssl != NULL) {
+    if (conn->flags & NSF_SSL_HANDSHAKE_DONE) {
+      n = SSL_read(conn->ssl, buf, sizeof(buf));
+    } else {
+      if (SSL_accept(conn->ssl) == 1) {
+        conn->flags |= NSF_SSL_HANDSHAKE_DONE;
+      }
+      return;
+    }
+  } else
+#endif
+  {
+    n = recv(conn->sock, buf, sizeof(buf), 0);
+  }
+
+#ifdef NS_ENABLE_HEXDUMP
+  hexdump(conn, buf, n, "<-");
+#endif
+
+  DBG(("%p <- %d bytes [%.*s%s]",
+       conn, n, n < 40 ? n : 40, buf, n < 40 ? "" : "..."));
+
+  if (ns_is_error(n)) {
+    conn->flags |= NSF_CLOSE_IMMEDIATELY;
+  } else if (n > 0) {
+    iobuf_append(&conn->recv_iobuf, buf, n);
+    call_user(conn, NS_RECV, &n);
+  }
+}
+
+static void write_to_socket(struct ns_connection *conn) {
+  struct iobuf *io = &conn->send_iobuf;
+  int n = 0;
+
+#ifdef NS_ENABLE_SSL
+  if (conn->ssl != NULL) {
+    n = SSL_write(conn->ssl, io->buf, io->len);
+  } else
+#endif
+  { n = send(conn->sock, io->buf, io->len, 0); }
+
+
+#ifdef NS_ENABLE_HEXDUMP
+  hexdump(conn, io->buf, n, "->");
+#endif
+
+  DBG(("%p -> %d bytes [%.*s%s]", conn, n, io->len < 40 ? io->len : 40,
+       io->buf, io->len < 40 ? "" : "..."));
+
+  if (ns_is_error(n)) {
+    conn->flags |= NSF_CLOSE_IMMEDIATELY;
+  } else if (n > 0) {
+    iobuf_remove(io, n);
+    //conn->num_bytes_sent += n;
+  }
+
+  if (io->len == 0 && conn->flags & NSF_FINISHED_SENDING_DATA) {
+    conn->flags |= NSF_CLOSE_IMMEDIATELY;
+  }
+
+  call_user(conn, NS_SEND, NULL);
+}
+
+int ns_send(struct ns_connection *conn, const void *buf, int len) {
+  return iobuf_append(&conn->send_iobuf, buf, len);
+}
+
+static void add_to_set(sock_t sock, fd_set *set, sock_t *max_fd) {
+  if (sock >= 0) FD_SET(sock, set);
+  if (sock > *max_fd) {
+    *max_fd = sock;
+  }
+}
+
+int ns_server_poll(struct ns_server *server, int milli) {
+  struct ns_connection *conn, *tmp_conn;
+  struct timeval tv;
+  fd_set read_set, write_set;
+  int num_active_connections = 0;
+  sock_t max_fd = INVALID_SOCKET;
+  time_t current_time = time(NULL);
+
+  if (server->listening_sock == INVALID_SOCKET &&
+      server->active_connections == NULL) return 0;
+
+  FD_ZERO(&read_set);
+  FD_ZERO(&write_set);
+  add_to_set(server->listening_sock, &read_set, &max_fd);
+
+  for (conn = server->active_connections; conn != NULL; conn = tmp_conn) {
+    tmp_conn = conn->next;
+    call_user(conn, NS_POLL, &current_time);
+    add_to_set(conn->sock, &read_set, &max_fd);
+    if (conn->flags & NSF_CONNECTING) {
+      add_to_set(conn->sock, &write_set, &max_fd);
+    }
+    if (conn->send_iobuf.len > 0 && !(conn->flags & NSF_BUFFER_BUT_DONT_SEND)) {
+      add_to_set(conn->sock, &write_set, &max_fd);
+    } else if (conn->flags & NSF_CLOSE_IMMEDIATELY) {
+      close_conn(conn);
+    }
+  }
+
+  tv.tv_sec = milli / 1000;
+  tv.tv_usec = (milli % 1000) * 1000;
+
+  if (select((int) max_fd + 1, &read_set, &write_set, NULL, &tv) > 0) {
+    // Accept new connections
+    if (server->listening_sock >= 0 &&
+        FD_ISSET(server->listening_sock, &read_set)) {
+      // We're not looping here, and accepting just one connection at
+      // a time. The reason is that eCos does not respect non-blocking
+      // flag on a listening socket and hangs in a loop.
+      if ((conn = accept_conn(server)) != NULL) {
+        conn->last_io_time = current_time;
+      }
+    }
+
+    for (conn = server->active_connections; conn != NULL; conn = tmp_conn) {
+      tmp_conn = conn->next;
+      if (FD_ISSET(conn->sock, &read_set)) {
+        conn->last_io_time = current_time;
+        read_from_socket(conn);
+      }
+      if (FD_ISSET(conn->sock, &write_set)) {
+        if (conn->flags & NSF_CONNECTING) {
+          read_from_socket(conn);
+        } else if (!(conn->flags & NSF_BUFFER_BUT_DONT_SEND)) {
+          conn->last_io_time = current_time;
+          write_to_socket(conn);
+        }
+      }
+    }
+  }
+
+  for (conn = server->active_connections; conn != NULL; conn = tmp_conn) {
+    tmp_conn = conn->next;
+    num_active_connections++;
+    if (conn->flags & NSF_CLOSE_IMMEDIATELY) {
+      close_conn(conn);
+    }
+  }
+  //DBG(("%d active connections", num_active_connections));
+
+  return num_active_connections;
+}
+
+struct ns_connection *ns_connect(struct ns_server *server, const char *host,
+                                 int port, int use_ssl, void *param) {
+  sock_t sock = INVALID_SOCKET;
+  struct sockaddr_in sin;
+  struct hostent *he = NULL;
+  struct ns_connection *conn = NULL;
+  int connect_ret_val;
+
+#ifndef NS_ENABLE_SSL
+  if (use_ssl) return 0;
+#endif
+
+  if (host == NULL || (he = gethostbyname(host)) == NULL ||
+      (sock = socket(PF_INET, SOCK_STREAM, 0)) == INVALID_SOCKET) {
+    DBG(("gethostbyname(%s) failed: %s", host, strerror(errno)));
+    return NULL;
+  }
+
+  sin.sin_family = AF_INET;
+  sin.sin_port = htons((uint16_t) port);
+  sin.sin_addr = * (struct in_addr *) he->h_addr_list[0];
+  set_non_blocking_mode(sock);
+
+  connect_ret_val = connect(sock, (struct sockaddr *) &sin, sizeof(sin));
+  if (ns_is_error(connect_ret_val)) {
+    return NULL;
+  } else if ((conn = (struct ns_connection *)
+              NS_MALLOC(sizeof(*conn))) == NULL) {
+    closesocket(sock);
+    return NULL;
+  }
+
+  memset(conn, 0, sizeof(*conn));
+  conn->server = server;
+  conn->sock = sock;
+  conn->connection_data = param;
+  conn->flags = NSF_CONNECTING;
+
+#ifdef NS_ENABLE_SSL
+  if (use_ssl &&
+      (conn->ssl = SSL_new(server->client_ssl_ctx)) != NULL) {
+    SSL_set_fd(conn->ssl, sock);
+  }
+#endif
+
+  add_connection(server, conn);
+  DBG(("%p %s:%d %d %p", conn, host, port, conn->sock, conn->ssl));
+
+  return conn;
+}
+
+struct ns_connection *ns_add_sock(struct ns_server *s, sock_t sock, void *p) {
+  struct ns_connection *conn;
+  if ((conn = (struct ns_connection *) NS_MALLOC(sizeof(*conn))) != NULL) {
+    memset(conn, 0, sizeof(*conn));
+    set_non_blocking_mode(sock);
+    conn->sock = sock;
+    conn->connection_data = p;
+    conn->server = s;
+    add_connection(s, conn);
+    DBG(("%p %d", conn, sock));
+  }
+  return conn;
+}
+
+void ns_iterate(struct ns_server *server, ns_callback_t cb, void *param) {
+  struct ns_connection *conn, *tmp_conn;
+
+  for (conn = server->active_connections; conn != NULL; conn = tmp_conn) {
+    tmp_conn = conn->next;
+    cb(conn, NS_POLL, param);
+  }
+}
+
+void ns_server_init(struct ns_server *s, void *server_data, ns_callback_t cb) {
+  memset(s, 0, sizeof(*s));
+  s->listening_sock = INVALID_SOCKET;
+  s->server_data = server_data;
+  s->callback = cb;
+
+#ifdef _WIN32
+  { WSADATA data; WSAStartup(MAKEWORD(2, 2), &data); }
+#else
+  // Ignore SIGPIPE signal, so if client cancels the request, it
+  // won't kill the whole process.
+  signal(SIGPIPE, SIG_IGN);
+#endif
+
+#ifdef NS_ENABLE_SSL
+  SSL_library_init();
+  s->client_ssl_ctx = SSL_CTX_new(SSLv23_client_method());
+#endif
+}
+
+void ns_server_free(struct ns_server *s) {
+  struct ns_connection *conn, *tmp_conn;
+
+  DBG(("%p", s));
+  if (s == NULL) return;
+  // Do one last poll, see https://github.com/cesanta/mongoose/issues/286
+  ns_server_poll(s, 0);
+
+  if (s->listening_sock != INVALID_SOCKET) {
+    closesocket(s->listening_sock);
+  }
+
+  for (conn = s->active_connections; conn != NULL; conn = tmp_conn) {
+    tmp_conn = conn->next;
+    close_conn(conn);
+  }
+
+#ifndef NS_DISABLE_SOCKETPAIR
+  //closesocket(s->ctl[0]);
+  //closesocket(s->ctl[1]);
+#endif
+
+#ifdef NS_ENABLE_SSL
+  if (s->ssl_ctx != NULL) SSL_CTX_free(s->ssl_ctx);
+  if (s->client_ssl_ctx != NULL) SSL_CTX_free(s->client_ssl_ctx);
+#endif
+}
+
 // net_skeleton end
 
 #include <ctype.h>