Split net/ from mmo/

10 files changed, 1971 insertions, 0 deletions

diff --git a/src/net/fwd.hpp b/src/net/fwd.hpp
new file mode 100644
index 0000000..ca54e8d
--- /dev/null
+++ b/src/net/fwd.hpp
@@ -0,0 +1,30 @@
+#ifndef TMWA_NET_FWD_HPP
+#define TMWA_NET_FWD_HPP
+//    net/fwd.hpp - list of type names for net lib
+//
+//    Copyright © 2014 Ben Longbons <b.r.longbons@gmail.com>
+//
+//    This file is part of The Mana World (Athena server)
+//
+//    This program is free software: you can redistribute it and/or modify
+//    it under the terms of the GNU General Public License as published by
+//    the Free Software Foundation, either version 3 of the License, or
+//    (at your option) any later version.
+//
+//    This program is distributed in the hope that it will be useful,
+//    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.
+//
+//    You should have received a copy of the GNU General Public License
+//    along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+# include "../sanity.hpp"
+
+class Session;
+
+class IP4Address;
+
+class TimerData;
+
+#endif // TMWA_NET_FWD_HPP
diff --git a/src/net/ip.cpp b/src/net/ip.cpp
new file mode 100644
index 0000000..591faab
--- /dev/null
+++ b/src/net/ip.cpp
@@ -0,0 +1,116 @@
+#include "ip.hpp"
+//    ip.cpp - Implementation of IP address functions.
+//
+//    Copyright © 2013 Ben Longbons <b.r.longbons@gmail.com>
+//
+//    This file is part of The Mana World (Athena server)
+//
+//    This program is free software: you can redistribute it and/or modify
+//    it under the terms of the GNU General Public License as published by
+//    the Free Software Foundation, either version 3 of the License, or
+//    (at your option) any later version.
+//
+//    This program is distributed in the hope that it will be useful,
+//    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.
+//
+//    You should have received a copy of the GNU General Public License
+//    along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+#include "../strings/xstring.hpp"
+#include "../strings/vstring.hpp"
+
+#include "../io/cxxstdio.hpp"
+
+#include "../mmo/extract.hpp"
+
+#include "../poison.hpp"
+
+bool extract(XString str, IP4Address *rv)
+{
+    if (str.endswith('.'))
+        return false;
+    uint8_t buf[4];
+    if (extract(str, record<'.'>(&buf[0], &buf[1], &buf[2], &buf[3])))
+    {
+        *rv = IP4Address(buf);
+        return true;
+    }
+    return false;
+}
+
+bool extract(XString str, IP4Mask *rv)
+{
+    IP4Address a, m;
+    unsigned b;
+    XString l, r;
+    if (str.endswith('/'))
+        return false;
+    if (extract(str, record<'/'>(&l, &r)))
+    {
+        // a.b.c.d/e.f.g.h or a.b.c.d/n
+        if (!extract(l, &a))
+            return false;
+        if (extract(r, &m))
+        {
+            *rv = IP4Mask(a, m);
+            return true;
+        }
+        if (!extract(r, &b) || b > 32)
+            return false;
+    }
+    else
+    {
+        // a.b.c.d or a.b.c.d. or a.b.c. or a.b. or a.
+        if (extract(str, &a))
+        {
+            *rv = IP4Mask(a, IP4_BROADCAST);
+            return true;
+        }
+        if (!str.endswith('.'))
+            return false;
+        uint8_t d[4] {};
+        if (extract(str, record<'.'>(&d[0], &d[1], &d[2], &d[3])))
+            b = 32;
+        else if (extract(str, record<'.'>(&d[0], &d[1], &d[2])))
+            b = 24;
+        else if (extract(str, record<'.'>(&d[0], &d[1])))
+            b = 16;
+        else if (extract(str, record<'.'>(&d[0])))
+            b = 8;
+        else
+            return false;
+        a = IP4Address(d);
+    }
+    // a is set; need to construct m from b
+    if (b == 0)
+        m = IP4Address();
+    else if (b == 32)
+        m = IP4_BROADCAST;
+    else
+    {
+        uint32_t s = -1;
+        s <<= (32 - b);
+        m = IP4Address({
+                static_cast<uint8_t>(s >> 24),
+                static_cast<uint8_t>(s >> 16),
+                static_cast<uint8_t>(s >> 8),
+                static_cast<uint8_t>(s >> 0),
+        });
+    }
+    *rv = IP4Mask(a, m);
+    return true;
+}
+
+VString<15> convert_for_printf(IP4Address a_)
+{
+    const uint8_t *a = a_.bytes();
+    return STRNPRINTF(16, "%hhu.%hhu.%hhu.%hhu"_fmt, a[0], a[1], a[2], a[3]);
+}
+
+VString<31> convert_for_printf(IP4Mask a)
+{
+    return STRNPRINTF(32, "%s/%s"_fmt,
+            a.addr(), a.mask());
+}
diff --git a/src/net/ip.hpp b/src/net/ip.hpp
new file mode 100644
index 0000000..42a4689
--- /dev/null
+++ b/src/net/ip.hpp
@@ -0,0 +1,165 @@
+#ifndef TMWA_NET_IP_HPP
+#define TMWA_NET_IP_HPP
+//    ip.hpp - classes to deal with IP addresses.
+//
+//    Copyright © 2013 Ben Longbons <b.r.longbons@gmail.com>
+//
+//    This file is part of The Mana World (Athena server)
+//
+//    This program is free software: you can redistribute it and/or modify
+//    it under the terms of the GNU General Public License as published by
+//    the Free Software Foundation, either version 3 of the License, or
+//    (at your option) any later version.
+//
+//    This program is distributed in the hope that it will be useful,
+//    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.
+//
+//    You should have received a copy of the GNU General Public License
+//    along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+# include "fwd.hpp"
+
+# include <netinet/in.h>
+
+# include <cstddef>
+# include <cstdint>
+
+# include "../strings/fwd.hpp"
+
+// TODO - in the long run ports belong here also
+// and of course, IPv6 stuff.
+// But what about unix socket addresses?
+
+/// Helper function
+template<class T, size_t n>
+constexpr
+bool _ce_a_lt(T (&a)[n], T (&b)[n], size_t i=0)
+{
+    return (i != n
+            && (a[i] < b[i]
+                || (a[i] == b[i]
+                    && _ce_a_lt(a, b, i + 1))));
+}
+
+/// A 32-bit Ipv4 address. Does not include a port.
+/// Guaranteed to be laid out like the network wants.
+class IP4Address
+{
+    uint8_t _addr[4];
+public:
+    constexpr
+    IP4Address()
+    : _addr{}
+    {}
+    constexpr explicit
+    IP4Address(const uint8_t (&a)[4])
+    : _addr{a[0], a[1], a[2], a[3]}
+    {}
+    explicit
+    IP4Address(struct in_addr addr)
+    {
+        static_assert(sizeof(addr) == sizeof(_addr), "4 bytes");
+        *this = IP4Address(reinterpret_cast<const uint8_t (&)[4]>(addr));
+    }
+    explicit
+    operator struct in_addr() const
+    {
+        return reinterpret_cast<const struct in_addr&>(_addr);
+    }
+
+    constexpr friend
+    IP4Address operator & (IP4Address l, IP4Address r)
+    {
+        return IP4Address({
+                static_cast<uint8_t>(l._addr[0] & r._addr[0]),
+                static_cast<uint8_t>(l._addr[1] & r._addr[1]),
+                static_cast<uint8_t>(l._addr[2] & r._addr[2]),
+                static_cast<uint8_t>(l._addr[3] & r._addr[3]),
+        });
+    }
+
+    IP4Address& operator &= (IP4Address m)
+    { return *this = *this & m; }
+
+    const uint8_t *bytes() const
+    { return _addr; }
+
+    constexpr friend
+    bool operator < (IP4Address l, IP4Address r)
+    {
+        return _ce_a_lt(l._addr, r._addr);
+    }
+
+    constexpr friend
+    bool operator > (IP4Address l, IP4Address r)
+    {
+        return _ce_a_lt(r._addr, l._addr);
+    }
+
+    constexpr friend
+    bool operator >= (IP4Address l, IP4Address r)
+    {
+        return !_ce_a_lt(l._addr, r._addr);
+    }
+
+    constexpr friend
+    bool operator <= (IP4Address l, IP4Address r)
+    {
+        return !_ce_a_lt(r._addr, l._addr);
+    }
+
+    constexpr friend
+    bool operator == (IP4Address l, IP4Address r)
+    {
+        return !(l < r || r < l);
+    }
+
+    constexpr friend
+    bool operator != (IP4Address l, IP4Address r)
+    {
+        return (l < r || r < l);
+    }
+};
+
+class IP4Mask
+{
+    IP4Address _addr, _mask;
+public:
+    constexpr
+    IP4Mask() : _addr(), _mask()
+    {}
+    constexpr
+    IP4Mask(IP4Address a, IP4Address m) : _addr(a & m), _mask(m)
+    {}
+
+    constexpr
+    IP4Address addr() const
+    { return _addr; }
+    constexpr
+    IP4Address mask() const
+    { return _mask; }
+
+    constexpr
+    bool covers(IP4Address a) const
+    {
+        return (a & _mask) == _addr;
+    }
+};
+
+
+constexpr
+IP4Address IP4_LOCALHOST({127, 0, 0, 1});
+constexpr
+IP4Address IP4_BROADCAST({255, 255, 255, 255});
+
+
+VString<15> convert_for_printf(IP4Address a);
+VString<31> convert_for_printf(IP4Mask m);
+
+bool extract(XString str, IP4Address *iv);
+
+bool extract(XString str, IP4Mask *iv);
+
+#endif // TMWA_NET_IP_HPP
diff --git a/src/net/ip.py b/src/net/ip.py
new file mode 100644
index 0000000..e6a8183
--- /dev/null
+++ b/src/net/ip.py
@@ -0,0 +1,14 @@
+class IP4Address(object):
+    ''' print an IP4Address
+    '''
+    __slots__ = ('_value')
+    name = 'IP4Address'
+    enabled = True
+
+    def __init__(self, value):
+        self._value = value
+
+    def to_string(self):
+        addr = self._value['_addr']
+        addr = tuple(int(addr[i]) for i in range(4))
+        return '%d.%d.%d.%d' % addr
diff --git a/src/net/ip_test.cpp b/src/net/ip_test.cpp
new file mode 100644
index 0000000..3c4d2ef
--- /dev/null
+++ b/src/net/ip_test.cpp
@@ -0,0 +1,355 @@
+#include "ip.hpp"
+//    ip_test.cpp - Testsuite for implementation of IP address functions.
+//
+//    Copyright © 2013 Ben Longbons <b.r.longbons@gmail.com>
+//
+//    This file is part of The Mana World (Athena server)
+//
+//    This program is free software: you can redistribute it and/or modify
+//    it under the terms of the GNU General Public License as published by
+//    the Free Software Foundation, either version 3 of the License, or
+//    (at your option) any later version.
+//
+//    This program is distributed in the hope that it will be useful,
+//    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.
+//
+//    You should have received a copy of the GNU General Public License
+//    along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+#include <gtest/gtest.h>
+
+#include "../strings/vstring.hpp"
+#include "../strings/literal.hpp"
+
+#include "../io/cxxstdio.hpp"
+
+#include "../poison.hpp"
+
+#define CB(X) (std::integral_constant<bool, (X)>::value)
+TEST(ip4addr, cmp)
+{
+    constexpr static
+    IP4Address a = IP4_LOCALHOST;
+    constexpr static
+    IP4Address b = IP4_BROADCAST;
+
+    EXPECT_FALSE(CB(a < a));
+    EXPECT_TRUE (CB(a < b));
+    EXPECT_FALSE(CB(b < a));
+    EXPECT_FALSE(CB(b < b));
+
+    EXPECT_FALSE(CB(a > a));
+    EXPECT_FALSE(CB(a > b));
+    EXPECT_TRUE (CB(b > a));
+    EXPECT_FALSE(CB(b > b));
+
+    EXPECT_TRUE (CB(a <= a));
+    EXPECT_TRUE (CB(a <= b));
+    EXPECT_FALSE(CB(b <= a));
+    EXPECT_TRUE (CB(b <= b));
+
+    EXPECT_TRUE (CB(a >= a));
+    EXPECT_FALSE(CB(a >= b));
+    EXPECT_TRUE (CB(b >= a));
+    EXPECT_TRUE (CB(b >= b));
+
+    EXPECT_TRUE (CB(a == a));
+    EXPECT_FALSE(CB(a == b));
+    EXPECT_FALSE(CB(b == a));
+    EXPECT_TRUE (CB(b == b));
+
+    EXPECT_FALSE(CB(a != a));
+    EXPECT_TRUE (CB(a != b));
+    EXPECT_TRUE (CB(b != a));
+    EXPECT_FALSE(CB(b != b));
+}
+
+TEST(ip4addr, str)
+{
+    IP4Address a;
+    EXPECT_EQ("0.0.0.0"_s, STRNPRINTF(17, "%s"_fmt, a));
+    EXPECT_EQ("127.0.0.1"_s, STRNPRINTF(17, "%s"_fmt, IP4_LOCALHOST));
+    EXPECT_EQ("255.255.255.255"_s, STRNPRINTF(17, "%s"_fmt, IP4_BROADCAST));
+}
+
+TEST(ip4addr, extract)
+{
+    IP4Address a;
+    EXPECT_TRUE(extract("0.0.0.0"_s, &a));
+    EXPECT_EQ("0.0.0.0"_s, STRNPRINTF(16, "%s"_fmt, a));
+    EXPECT_TRUE(extract("127.0.0.1"_s, &a));
+    EXPECT_EQ("127.0.0.1"_s, STRNPRINTF(16, "%s"_fmt, a));
+    EXPECT_TRUE(extract("255.255.255.255"_s, &a));
+    EXPECT_EQ("255.255.255.255"_s, STRNPRINTF(16, "%s"_fmt, a));
+    EXPECT_TRUE(extract("1.2.3.4"_s, &a));
+    EXPECT_EQ("1.2.3.4"_s, STRNPRINTF(16, "%s"_fmt, a));
+
+    EXPECT_FALSE(extract("1.2.3.4.5"_s, &a));
+    EXPECT_FALSE(extract("1.2.3.4."_s, &a));
+    EXPECT_FALSE(extract("1.2.3."_s, &a));
+    EXPECT_FALSE(extract("1.2.3"_s, &a));
+    EXPECT_FALSE(extract("1.2."_s, &a));
+    EXPECT_FALSE(extract("1.2"_s, &a));
+    EXPECT_FALSE(extract("1."_s, &a));
+    EXPECT_FALSE(extract("1"_s, &a));
+    EXPECT_FALSE(extract(""_s, &a));
+}
+
+
+TEST(ip4mask, body)
+{
+    IP4Mask m;
+    EXPECT_EQ(IP4Address(), m.addr());
+    EXPECT_EQ(IP4Address(), m.mask());
+    m = IP4Mask(IP4_LOCALHOST, IP4_BROADCAST);
+    EXPECT_EQ(IP4_LOCALHOST, m.addr());
+    EXPECT_EQ(IP4_BROADCAST, m.mask());
+}
+
+TEST(ip4mask, str)
+{
+    IP4Mask m;
+    EXPECT_EQ("0.0.0.0/0.0.0.0"_s, STRNPRINTF(33, "%s"_fmt, m));
+    m = IP4Mask(IP4_LOCALHOST, IP4_BROADCAST);
+    EXPECT_EQ("127.0.0.1/255.255.255.255"_s, STRNPRINTF(33, "%s"_fmt, m));
+}
+
+TEST(ip4mask, extract)
+{
+    IP4Mask m;
+    EXPECT_FALSE(extract("9.8.7.6/33"_s, &m));
+    EXPECT_FALSE(extract("9.8.7.6.5"_s, &m));
+    EXPECT_FALSE(extract("9.8.7.6/"_s, &m));
+    EXPECT_FALSE(extract("9.8.7"_s, &m));
+    EXPECT_FALSE(extract("9.8"_s, &m));
+    EXPECT_FALSE(extract("9"_s, &m));
+
+    EXPECT_TRUE(extract("127.0.0.1"_s, &m));
+    EXPECT_EQ("127.0.0.1/255.255.255.255"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("127.0.0.1."_s, &m));
+    EXPECT_EQ("127.0.0.1/255.255.255.255"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("127.0.0."_s, &m));
+    EXPECT_EQ("127.0.0.0/255.255.255.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("127.0."_s, &m));
+    EXPECT_EQ("127.0.0.0/255.255.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("127."_s, &m));
+    EXPECT_EQ("127.0.0.0/255.0.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+
+    EXPECT_TRUE(extract("1.2.3.4/255.255.255.255"_s, &m));
+    EXPECT_EQ("1.2.3.4/255.255.255.255"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("1.2.3.0/255.255.255.0"_s, &m));
+    EXPECT_EQ("1.2.3.0/255.255.255.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("1.2.0.4/255.255.0.255"_s, &m));
+    EXPECT_EQ("1.2.0.4/255.255.0.255"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("1.2.0.0/255.255.0.0"_s, &m));
+    EXPECT_EQ("1.2.0.0/255.255.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("1.0.3.4/255.0.255.255"_s, &m));
+    EXPECT_EQ("1.0.3.4/255.0.255.255"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("1.0.3.0/255.0.255.0"_s, &m));
+    EXPECT_EQ("1.0.3.0/255.0.255.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("1.0.0.4/255.0.0.255"_s, &m));
+    EXPECT_EQ("1.0.0.4/255.0.0.255"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("1.0.0.0/255.0.0.0"_s, &m));
+    EXPECT_EQ("1.0.0.0/255.0.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.2.3.4/0.255.255.255"_s, &m));
+    EXPECT_EQ("0.2.3.4/0.255.255.255"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.2.3.0/0.255.255.0"_s, &m));
+    EXPECT_EQ("0.2.3.0/0.255.255.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.2.0.4/0.255.0.255"_s, &m));
+    EXPECT_EQ("0.2.0.4/0.255.0.255"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.2.0.0/0.255.0.0"_s, &m));
+    EXPECT_EQ("0.2.0.0/0.255.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.3.4/0.0.255.255"_s, &m));
+    EXPECT_EQ("0.0.3.4/0.0.255.255"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.3.0/0.0.255.0"_s, &m));
+    EXPECT_EQ("0.0.3.0/0.0.255.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.4/0.0.0.255"_s, &m));
+    EXPECT_EQ("0.0.0.4/0.0.0.255"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/0.0.0.0"_s, &m));
+    EXPECT_EQ("0.0.0.0/0.0.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+
+    // please don't do this
+    EXPECT_TRUE(extract("120.248.200.217/89.57.126.5"_s, &m));
+    EXPECT_EQ("88.56.72.1/89.57.126.5"_s, STRNPRINTF(32, "%s"_fmt, m));
+
+    EXPECT_TRUE(extract("0.0.0.0/32"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.255.255.255"_s, STRNPRINTF(32, "%s"_fmt, m));
+
+    EXPECT_TRUE(extract("0.0.0.0/31"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.255.255.254"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/30"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.255.255.252"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/29"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.255.255.248"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/28"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.255.255.240"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/27"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.255.255.224"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/26"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.255.255.192"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/25"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.255.255.128"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/24"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.255.255.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+
+    EXPECT_TRUE(extract("0.0.0.0/23"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.255.254.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/22"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.255.252.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/21"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.255.248.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/20"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.255.240.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/19"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.255.224.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/18"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.255.192.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/17"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.255.128.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/16"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.255.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+
+    EXPECT_TRUE(extract("0.0.0.0/15"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.254.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/14"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.252.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/13"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.248.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/12"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.240.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/11"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.224.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/10"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.192.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/9"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.128.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/8"_s, &m));
+    EXPECT_EQ("0.0.0.0/255.0.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+
+    EXPECT_TRUE(extract("0.0.0.0/7"_s, &m));
+    EXPECT_EQ("0.0.0.0/254.0.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/6"_s, &m));
+    EXPECT_EQ("0.0.0.0/252.0.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/5"_s, &m));
+    EXPECT_EQ("0.0.0.0/248.0.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/4"_s, &m));
+    EXPECT_EQ("0.0.0.0/240.0.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/3"_s, &m));
+    EXPECT_EQ("0.0.0.0/224.0.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/2"_s, &m));
+    EXPECT_EQ("0.0.0.0/192.0.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/1"_s, &m));
+    EXPECT_EQ("0.0.0.0/128.0.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+    EXPECT_TRUE(extract("0.0.0.0/0"_s, &m));
+    EXPECT_EQ("0.0.0.0/0.0.0.0"_s, STRNPRINTF(32, "%s"_fmt, m));
+}
+
+TEST(ip4mask, cover)
+{
+    IP4Address a;
+    IP4Address b = IP4_BROADCAST;
+    IP4Address l = IP4_LOCALHOST;
+    IP4Address h({127, 255, 255, 255});
+    IP4Address p24l({10, 0, 0, 0});
+    IP4Address p24h({10, 255, 255, 255});
+    IP4Address p20l({172, 16, 0, 0});
+    IP4Address p20h({172, 31, 255, 255});
+    IP4Address p16l({192, 168, 0, 0});
+    IP4Address p16h({192, 168, 255, 255});
+    IP4Mask m;
+    EXPECT_TRUE(m.covers(a));
+    EXPECT_TRUE(m.covers(b));
+    EXPECT_TRUE(m.covers(l));
+    EXPECT_TRUE(m.covers(h));
+    EXPECT_TRUE(m.covers(p24l));
+    EXPECT_TRUE(m.covers(p24h));
+    EXPECT_TRUE(m.covers(p20l));
+    EXPECT_TRUE(m.covers(p20h));
+    EXPECT_TRUE(m.covers(p16l));
+    EXPECT_TRUE(m.covers(p16h));
+    m = IP4Mask(l, a);
+    EXPECT_TRUE(m.covers(a));
+    EXPECT_TRUE(m.covers(b));
+    EXPECT_TRUE(m.covers(l));
+    EXPECT_TRUE(m.covers(h));
+    EXPECT_TRUE(m.covers(p24l));
+    EXPECT_TRUE(m.covers(p24h));
+    EXPECT_TRUE(m.covers(p20l));
+    EXPECT_TRUE(m.covers(p20h));
+    EXPECT_TRUE(m.covers(p16l));
+    EXPECT_TRUE(m.covers(p16h));
+    m = IP4Mask(l, b);
+    EXPECT_FALSE(m.covers(a));
+    EXPECT_FALSE(m.covers(b));
+    EXPECT_TRUE(m.covers(l));
+    EXPECT_FALSE(m.covers(h));
+    EXPECT_FALSE(m.covers(p24l));
+    EXPECT_FALSE(m.covers(p24h));
+    EXPECT_FALSE(m.covers(p20l));
+    EXPECT_FALSE(m.covers(p20h));
+    EXPECT_FALSE(m.covers(p16l));
+    EXPECT_FALSE(m.covers(p16h));
+
+    // but the really useful ones are with partial masks
+    m = IP4Mask(IP4Address({10, 0, 0, 0}), IP4Address({255, 0, 0, 0}));
+    EXPECT_FALSE(m.covers(a));
+    EXPECT_FALSE(m.covers(b));
+    EXPECT_FALSE(m.covers(l));
+    EXPECT_FALSE(m.covers(h));
+    EXPECT_TRUE(m.covers(p24l));
+    EXPECT_TRUE(m.covers(p24h));
+    EXPECT_FALSE(m.covers(p20l));
+    EXPECT_FALSE(m.covers(p20h));
+    EXPECT_FALSE(m.covers(p16l));
+    EXPECT_FALSE(m.covers(p16h));
+    EXPECT_FALSE(m.covers(IP4Address({9, 255, 255, 255})));
+    EXPECT_FALSE(m.covers(IP4Address({11, 0, 0, 0})));
+    m = IP4Mask(IP4Address({127, 0, 0, 0}), IP4Address({255, 0, 0, 0}));
+    EXPECT_FALSE(m.covers(a));
+    EXPECT_FALSE(m.covers(b));
+    EXPECT_TRUE(m.covers(l));
+    EXPECT_TRUE(m.covers(h));
+    EXPECT_FALSE(m.covers(p24l));
+    EXPECT_FALSE(m.covers(p24h));
+    EXPECT_FALSE(m.covers(p20l));
+    EXPECT_FALSE(m.covers(p20h));
+    EXPECT_FALSE(m.covers(p16l));
+    EXPECT_FALSE(m.covers(p16h));
+    EXPECT_FALSE(m.covers(IP4Address({126, 255, 255, 255})));
+    EXPECT_FALSE(m.covers(IP4Address({128, 0, 0, 0})));
+    m = IP4Mask(IP4Address({172, 16, 0, 0}), IP4Address({255, 240, 0, 0}));
+    EXPECT_FALSE(m.covers(a));
+    EXPECT_FALSE(m.covers(b));
+    EXPECT_FALSE(m.covers(l));
+    EXPECT_FALSE(m.covers(h));
+    EXPECT_FALSE(m.covers(p24l));
+    EXPECT_FALSE(m.covers(p24h));
+    EXPECT_TRUE(m.covers(p20l));
+    EXPECT_TRUE(m.covers(p20h));
+    EXPECT_FALSE(m.covers(p16l));
+    EXPECT_FALSE(m.covers(p16h));
+    EXPECT_FALSE(m.covers(IP4Address({172, 15, 255, 255})));
+    EXPECT_FALSE(m.covers(IP4Address({172, 32, 0, 0})));
+    m = IP4Mask(IP4Address({192, 168, 0, 0}), IP4Address({255, 255, 0, 0}));
+    EXPECT_FALSE(m.covers(a));
+    EXPECT_FALSE(m.covers(b));
+    EXPECT_FALSE(m.covers(l));
+    EXPECT_FALSE(m.covers(h));
+    EXPECT_FALSE(m.covers(p24l));
+    EXPECT_FALSE(m.covers(p24h));
+    EXPECT_FALSE(m.covers(p20l));
+    EXPECT_FALSE(m.covers(p20h));
+    EXPECT_TRUE(m.covers(p16l));
+    EXPECT_TRUE(m.covers(p16h));
+    EXPECT_FALSE(m.covers(IP4Address({192, 167, 255, 255})));
+    EXPECT_FALSE(m.covers(IP4Address({192, 169, 0, 0})));
+
+    // OTOH this is crazy
+    EXPECT_TRUE(extract("120.248.200.217/89.57.126.5"_s, &m));
+    EXPECT_TRUE(m.covers(IP4Address({120, 248, 200, 217})));
+    EXPECT_TRUE(m.covers(IP4Address({88, 56, 72, 1})));
+    EXPECT_FALSE(m.covers(IP4Address({88, 56, 72, 0})));
+    EXPECT_FALSE(m.covers(IP4Address({88, 56, 72, 255})));
+}
diff --git a/src/net/socket.cpp b/src/net/socket.cpp
new file mode 100644
index 0000000..6880bfa
--- /dev/null
+++ b/src/net/socket.cpp
@@ -0,0 +1,511 @@
+#include "socket.hpp"
+//    socket.cpp - Network event system.
+//
+//    Copyright © ????-2004 Athena Dev Teams
+//    Copyright © 2004-2011 The Mana World Development Team
+//    Copyright © 2011-2014 Ben Longbons <b.r.longbons@gmail.com>
+//    Copyright © 2013 MadCamel
+//
+//    This file is part of The Mana World (Athena server)
+//
+//    This program is free software: you can redistribute it and/or modify
+//    it under the terms of the GNU General Public License as published by
+//    the Free Software Foundation, either version 3 of the License, or
+//    (at your option) any later version.
+//
+//    This program is distributed in the hope that it will be useful,
+//    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.
+//
+//    You should have received a copy of the GNU General Public License
+//    along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+#include <netinet/tcp.h>
+#include <sys/socket.h>
+
+#include <fcntl.h>
+
+#include <cstdlib>
+
+#include <array>
+
+#include "../compat/memory.hpp"
+
+#include "../io/cxxstdio.hpp"
+
+// TODO get rid of ordering violations
+#include "../mmo/utils.hpp"
+#include "../mmo/core.hpp"
+
+#include "timer.hpp"
+
+#include "../poison.hpp"
+
+static
+io::FD_Set readfds;
+static
+int fd_max;
+
+static
+const uint32_t RFIFO_SIZE = 65536;
+static
+const uint32_t WFIFO_SIZE = 65536;
+
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wold-style-cast"
+static
+std::array<std::unique_ptr<Session>, FD_SETSIZE> session;
+#pragma GCC diagnostic pop
+
+Session::Session(SessionIO io, SessionParsers p)
+: created()
+, connected()
+, eof()
+, timed_close()
+, rdata(), wdata()
+, max_rdata(), max_wdata()
+, rdata_size(), wdata_size()
+, rdata_pos()
+, client_ip()
+, func_recv()
+, func_send()
+, func_parse()
+, func_delete()
+, for_inferior()
+, session_data()
+, fd()
+{
+    set_io(io);
+    set_parsers(p);
+}
+void Session::set_io(SessionIO io)
+{
+    func_send = io.func_send;
+    func_recv = io.func_recv;
+}
+void Session::set_parsers(SessionParsers p)
+{
+    func_parse = p.func_parse;
+    func_delete = p.func_delete;
+}
+
+
+void set_session(io::FD fd, std::unique_ptr<Session> sess)
+{
+    int f = fd.uncast_dammit();
+    assert (0 <= f && f < FD_SETSIZE);
+    session[f] = std::move(sess);
+}
+Session *get_session(io::FD fd)
+{
+    int f = fd.uncast_dammit();
+    if (0 <= f && f < FD_SETSIZE)
+        return session[f].get();
+    return nullptr;
+}
+void reset_session(io::FD fd)
+{
+    int f = fd.uncast_dammit();
+    assert (0 <= f && f < FD_SETSIZE);
+    session[f] = nullptr;
+}
+int get_fd_max() { return fd_max; }
+IteratorPair<ValueIterator<io::FD, IncrFD>> iter_fds()
+{
+    return {io::FD::cast_dammit(0), io::FD::cast_dammit(fd_max)};
+}
+
+/// clean up by discarding handled bytes
+inline
+void RFIFOFLUSH(Session *s)
+{
+    really_memmove(&s->rdata[0], &s->rdata[s->rdata_pos], RFIFOREST(s));
+    s->rdata_size = RFIFOREST(s);
+    s->rdata_pos = 0;
+}
+
+/// how much room there is to read more data
+inline
+size_t RFIFOSPACE(Session *s)
+{
+    return s->max_rdata - s->rdata_size;
+}
+
+
+/// Read from socket to the queue
+static
+void recv_to_fifo(Session *s)
+{
+    ssize_t len = s->fd.read(&s->rdata[s->rdata_size],
+                        RFIFOSPACE(s));
+
+    if (len > 0)
+    {
+        s->rdata_size += len;
+        s->connected = 1;
+    }
+    else
+    {
+        s->set_eof();
+    }
+}
+
+static
+void send_from_fifo(Session *s)
+{
+    ssize_t len = s->fd.write(&s->wdata[0], s->wdata_size);
+
+    if (len > 0)
+    {
+        s->wdata_size -= len;
+        if (s->wdata_size)
+        {
+            really_memmove(&s->wdata[0], &s->wdata[len],
+                     s->wdata_size);
+        }
+        s->connected = 1;
+    }
+    else
+    {
+        s->set_eof();
+    }
+}
+
+static
+void nothing_delete(Session *s)
+{
+    (void)s;
+}
+
+static
+void connect_client(Session *ls)
+{
+    struct sockaddr_in client_address;
+    socklen_t len = sizeof(client_address);
+
+    io::FD fd = ls->fd.accept(reinterpret_cast<struct sockaddr *>(&client_address), &len);
+    if (fd == io::FD())
+    {
+        perror("accept");
+        return;
+    }
+    if (fd.uncast_dammit() >= SOFT_LIMIT)
+    {
+        FPRINTF(stderr, "softlimit reached, disconnecting : %d\n"_fmt, fd.uncast_dammit());
+        fd.shutdown(SHUT_RDWR);
+        fd.close();
+        return;
+    }
+    if (fd_max <= fd.uncast_dammit())
+    {
+        fd_max = fd.uncast_dammit() + 1;
+    }
+
+    const int yes = 1;
+    /// Allow to bind() again after the server restarts.
+    // Since the socket is still in the TIME_WAIT, there's a possibility
+    // that formerly lost packets might be delivered and confuse the server.
+    fd.setsockopt(SOL_SOCKET, SO_REUSEADDR, &yes, sizeof yes);
+    /// Send packets as soon as possible
+    /// even if the kernel thinks there is too little for it to be worth it!
+    /// Testing shows this is indeed a good idea.
+    fd.setsockopt(IPPROTO_TCP, TCP_NODELAY, &yes, sizeof yes);
+
+    // Linux-ism: Set socket options to optimize for thin streams
+    // See http://lwn.net/Articles/308919/ and
+    // Documentation/networking/tcp-thin.txt .. Kernel 3.2+
+#ifdef TCP_THIN_LINEAR_TIMEOUTS
+    fd.setsockopt(IPPROTO_TCP, TCP_THIN_LINEAR_TIMEOUTS, &yes, sizeof yes);
+#endif
+#ifdef TCP_THIN_DUPACK
+    fd.setsockopt(IPPROTO_TCP, TCP_THIN_DUPACK, &yes, sizeof yes);
+#endif
+
+    readfds.set(fd);
+
+    fd.fcntl(F_SETFL, O_NONBLOCK);
+
+    set_session(fd, make_unique<Session>(
+                SessionIO{.func_recv= recv_to_fifo, .func_send= send_from_fifo},
+                ls->for_inferior));
+    Session *s = get_session(fd);
+    s->fd = fd;
+    s->rdata.new_(RFIFO_SIZE);
+    s->wdata.new_(WFIFO_SIZE);
+    s->max_rdata = RFIFO_SIZE;
+    s->max_wdata = WFIFO_SIZE;
+    s->client_ip = IP4Address(client_address.sin_addr);
+    s->created = TimeT::now();
+    s->connected = 0;
+}
+
+Session *make_listen_port(uint16_t port, SessionParsers inferior)
+{
+    struct sockaddr_in server_address;
+    io::FD fd = io::FD::socket(AF_INET, SOCK_STREAM, 0);
+    if (fd == io::FD())
+    {
+        perror("socket");
+        return nullptr;
+    }
+    if (fd_max <= fd.uncast_dammit())
+        fd_max = fd.uncast_dammit() + 1;
+
+    fd.fcntl(F_SETFL, O_NONBLOCK);
+
+    const int yes = 1;
+    /// Allow to bind() again after the server restarts.
+    // Since the socket is still in the TIME_WAIT, there's a possibility
+    // that formerly lost packets might be delivered and confuse the server.
+    fd.setsockopt(SOL_SOCKET, SO_REUSEADDR, &yes, sizeof yes);
+    /// Send packets as soon as possible
+    /// even if the kernel thinks there is too little for it to be worth it!
+    // I'm not convinced this is a good idea; although in minimizes the
+    // latency for an individual write, it increases traffic in general.
+    fd.setsockopt(IPPROTO_TCP, TCP_NODELAY, &yes, sizeof yes);
+
+    server_address.sin_family = AF_INET;
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wold-style-cast"
+#if __GNUC__ > 4 || __GNUC_MINOR__ >= 8
+# pragma GCC diagnostic ignored "-Wuseless-cast"
+#endif
+    server_address.sin_addr.s_addr = htonl(INADDR_ANY);
+    server_address.sin_port = htons(port);
+#pragma GCC diagnostic pop
+
+    if (fd.bind(reinterpret_cast<struct sockaddr *>(&server_address),
+              sizeof(server_address)) == -1)
+    {
+        perror("bind");
+        exit(1);
+    }
+    if (fd.listen(5) == -1)
+    {                           /* error */
+        perror("listen");
+        exit(1);
+    }
+
+    readfds.set(fd);
+
+    set_session(fd, make_unique<Session>(
+                SessionIO{.func_recv= connect_client, .func_send= nullptr},
+                SessionParsers{.func_parse= nullptr, .func_delete= nothing_delete}));
+    Session *s = get_session(fd);
+    s->for_inferior = inferior;
+    s->fd = fd;
+
+    s->created = TimeT::now();
+    s->connected = 1;
+
+    return s;
+}
+
+Session *make_connection(IP4Address ip, uint16_t port, SessionParsers parsers)
+{
+    struct sockaddr_in server_address;
+    io::FD fd = io::FD::socket(AF_INET, SOCK_STREAM, 0);
+    if (fd == io::FD())
+    {
+        perror("socket");
+        return nullptr;
+    }
+    if (fd_max <= fd.uncast_dammit())
+        fd_max = fd.uncast_dammit() + 1;
+
+    const int yes = 1;
+    /// Allow to bind() again after the server restarts.
+    // Since the socket is still in the TIME_WAIT, there's a possibility
+    // that formerly lost packets might be delivered and confuse the server.
+    fd.setsockopt(SOL_SOCKET, SO_REUSEADDR, &yes, sizeof yes);
+    /// Send packets as soon as possible
+    /// even if the kernel thinks there is too little for it to be worth it!
+    // I'm not convinced this is a good idea; although in minimizes the
+    // latency for an individual write, it increases traffic in general.
+    fd.setsockopt(IPPROTO_TCP, TCP_NODELAY, &yes, sizeof yes);
+
+    server_address.sin_family = AF_INET;
+    server_address.sin_addr = in_addr(ip);
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wold-style-cast"
+#if __GNUC__ > 4 || __GNUC_MINOR__ >= 8
+# pragma GCC diagnostic ignored "-Wuseless-cast"
+#endif
+    server_address.sin_port = htons(port);
+#pragma GCC diagnostic pop
+
+    fd.fcntl(F_SETFL, O_NONBLOCK);
+
+    /// Errors not caught - we must not block
+    /// Let the main select() loop detect when we know the state
+    fd.connect(reinterpret_cast<struct sockaddr *>(&server_address),
+             sizeof(struct sockaddr_in));
+
+    readfds.set(fd);
+
+    set_session(fd, make_unique<Session>(
+                SessionIO{.func_recv= recv_to_fifo, .func_send= send_from_fifo},
+                parsers));
+    Session *s = get_session(fd);
+    s->fd = fd;
+    s->rdata.new_(RFIFO_SIZE);
+    s->wdata.new_(WFIFO_SIZE);
+
+    s->max_rdata = RFIFO_SIZE;
+    s->max_wdata = WFIFO_SIZE;
+    s->created = TimeT::now();
+    s->connected = 1;
+
+    return s;
+}
+
+void delete_session(Session *s)
+{
+    if (!s)
+        return;
+    // this needs to be before the fd_max--
+    s->func_delete(s);
+
+    io::FD fd = s->fd;
+    // If this was the highest fd, decrease it
+    // We could add a loop to decrement fd_max further for every null session,
+    // but this is cheap and good enough for the typical case
+    if (fd.uncast_dammit() == fd_max - 1)
+        fd_max--;
+    readfds.clr(fd);
+    {
+        s->rdata.delete_();
+        s->wdata.delete_();
+        s->session_data.reset();
+        reset_session(fd);
+    }
+
+    // just close() would try to keep sending buffers
+    fd.shutdown(SHUT_RDWR);
+    fd.close();
+}
+
+void realloc_fifo(Session *s, size_t rfifo_size, size_t wfifo_size)
+{
+    if (s->max_rdata != rfifo_size && s->rdata_size < rfifo_size)
+    {
+        s->rdata.resize(rfifo_size);
+        s->max_rdata = rfifo_size;
+    }
+    if (s->max_wdata != wfifo_size && s->wdata_size < wfifo_size)
+    {
+        s->wdata.resize(wfifo_size);
+        s->max_wdata = wfifo_size;
+    }
+}
+
+void WFIFOSET(Session *s, size_t len)
+{
+    if (s->wdata_size + len + 16384 > s->max_wdata)
+    {
+        realloc_fifo(s, s->max_rdata, s->max_wdata << 1);
+        PRINTF("socket: %d wdata expanded to %zu bytes.\n"_fmt, s, s->max_wdata);
+    }
+    if (s->wdata_size + len + 2048 < s->max_wdata)
+        s->wdata_size += len;
+    else
+        FPRINTF(stderr, "socket: %d wdata lost !!\n"_fmt, s), abort();
+}
+
+void do_sendrecv(interval_t next_ms)
+{
+    bool any = false;
+    io::FD_Set rfd = readfds, wfd;
+    for (io::FD i : iter_fds())
+    {
+        Session *s = get_session(i);
+        if (s)
+        {
+            any = true;
+            if (s->wdata_size)
+                wfd.set(i);
+        }
+    }
+    if (!any)
+    {
+        if (!has_timers())
+        {
+            PRINTF("Shutting down - nothing to do\n"_fmt);
+            // TODO hoist this
+            runflag = false;
+        }
+        return;
+    }
+    struct timeval timeout;
+    {
+        std::chrono::seconds next_s = std::chrono::duration_cast<std::chrono::seconds>(next_ms);
+        std::chrono::microseconds next_us = next_ms - next_s;
+        timeout.tv_sec = next_s.count();
+        timeout.tv_usec = next_us.count();
+    }
+    if (io::FD_Set::select(fd_max, &rfd, &wfd, NULL, &timeout) <= 0)
+        return;
+    for (io::FD i : iter_fds())
+    {
+        Session *s = get_session(i);
+        if (!s)
+            continue;
+        if (wfd.isset(i) && !s->eof)
+        {
+            if (s->func_send)
+                //send_from_fifo(i);
+                s->func_send(s);
+        }
+        if (rfd.isset(i) && !s->eof)
+        {
+            if (s->func_recv)
+                //recv_to_fifo(i);
+                //or connect_client(i);
+                s->func_recv(s);
+        }
+    }
+}
+
+void do_parsepacket(void)
+{
+    for (io::FD i : iter_fds())
+    {
+        Session *s = get_session(i);
+        if (!s)
+            continue;
+        if (!s->connected
+            && static_cast<time_t>(TimeT::now()) - static_cast<time_t>(s->created) > CONNECT_TIMEOUT)
+        {
+            PRINTF("Session #%d timed out\n"_fmt, s);
+            s->set_eof();
+        }
+        if (s->rdata_size && !s->eof && s->func_parse)
+        {
+            s->func_parse(s);
+            /// some func_parse may call delete_session
+            // (that's kind of evil)
+            s = get_session(i);
+            if (!s)
+                continue;
+        }
+        if (s->eof)
+        {
+            delete_session(s);
+            continue;
+        }
+        /// Reclaim buffer space for what was read
+        RFIFOFLUSH(s);
+    }
+}
+
+void RFIFOSKIP(Session *s, size_t len)
+{
+    s->rdata_pos += len;
+
+    if (s->rdata_size < s->rdata_pos)
+    {
+        FPRINTF(stderr, "too many skip\n"_fmt);
+        abort();
+    }
+}
diff --git a/src/net/socket.hpp b/src/net/socket.hpp
new file mode 100644
index 0000000..94198f6
--- /dev/null
+++ b/src/net/socket.hpp
@@ -0,0 +1,425 @@
+#ifndef TMWA_NET_SOCKET_HPP
+#define TMWA_NET_SOCKET_HPP
+//    socket.hpp - Network event system.
+//
+//    Copyright © ????-2004 Athena Dev Teams
+//    Copyright © 2004-2011 The Mana World Development Team
+//    Copyright © 2011-2014 Ben Longbons <b.r.longbons@gmail.com>
+//
+//    This file is part of The Mana World (Athena server)
+//
+//    This program is free software: you can redistribute it and/or modify
+//    it under the terms of the GNU General Public License as published by
+//    the Free Software Foundation, either version 3 of the License, or
+//    (at your option) any later version.
+//
+//    This program is distributed in the hope that it will be useful,
+//    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.
+//
+//    You should have received a copy of the GNU General Public License
+//    along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+# include "fwd.hpp"
+
+# include <algorithm>
+
+# include <sys/select.h>
+
+# include <memory>
+
+# include "../compat/iter.hpp"
+# include "../compat/rawmem.hpp"
+# include "../compat/time_t.hpp"
+
+# include "../strings/astring.hpp"
+# include "../strings/vstring.hpp"
+# include "../strings/xstring.hpp"
+
+# include "../generic/dumb_ptr.hpp"
+
+# include "../io/fd.hpp"
+
+# include "ip.hpp"
+# include "timer.t.hpp"
+
+struct SessionData
+{
+};
+struct SessionDeleter
+{
+    // defined per-server
+    void operator()(SessionData *sd);
+};
+
+struct SessionIO
+{
+    void (*func_recv)(Session *);
+    void (*func_send)(Session *);
+};
+
+struct SessionParsers
+{
+    void (*func_parse)(Session *);
+    void (*func_delete)(Session *);
+};
+
+struct Session
+{
+    Session(SessionIO, SessionParsers);
+    Session(Session&&) = delete;
+    Session& operator = (Session&&) = delete;
+
+    void set_io(SessionIO);
+    void set_parsers(SessionParsers);
+
+    /// Checks whether a newly-connected socket actually does anything
+    TimeT created;
+    bool connected;
+
+private:
+    /// Flag needed since structure must be freed in a server-dependent manner
+    bool eof;
+public:
+    void set_eof() { eof = true; }
+
+    /// Currently used by clif_setwaitclose
+    Timer timed_close;
+
+    /// Since this is a single-threaded application, it can't block
+    /// These are the read/write queues
+    dumb_ptr<uint8_t[]> rdata, wdata;
+    size_t max_rdata, max_wdata;
+    /// How much is actually in the queue
+    size_t rdata_size, wdata_size;
+    /// How much has already been read from the queue
+    /// Note that there is no need for a wdata_pos
+    size_t rdata_pos;
+
+    IP4Address client_ip;
+
+private:
+    /// Send or recieve
+    /// Only called when select() indicates the socket is ready
+    /// If, after that, nothing is read, it sets eof
+    // These could probably be hard-coded with a little work
+    void (*func_recv)(Session *);
+    void (*func_send)(Session *);
+    /// This is the important one
+    /// Set to different functions depending on whether the connection
+    /// is a player or a server/ladmin
+    void (*func_parse)(Session *);
+    /// Cleanup function since we're not fully RAII yet
+    void (*func_delete)(Session *);
+
+public:
+    // this really ought to be part of session_data, once that gets sane
+    SessionParsers for_inferior;
+
+    /// Server-specific data type
+    // (this really should include the deleter, but ...)
+    std::unique_ptr<SessionData, SessionDeleter> session_data;
+
+    io::FD fd;
+
+    friend void do_sendrecv(interval_t next);
+    friend void do_parsepacket(void);
+    friend void delete_session(Session *);
+};
+
+inline
+int convert_for_printf(Session *s)
+{
+    return s->fd.uncast_dammit();
+}
+
+// save file descriptors for important stuff
+constexpr int SOFT_LIMIT = FD_SETSIZE - 50;
+
+// socket timeout to establish a full connection in seconds
+constexpr int CONNECT_TIMEOUT = 15;
+
+
+void set_session(io::FD fd, std::unique_ptr<Session> sess);
+Session *get_session(io::FD fd);
+void reset_session(io::FD fd);
+int get_fd_max();
+
+class IncrFD
+{
+public:
+    static
+    io::FD inced(io::FD v)
+    {
+        return io::FD::cast_dammit(v.uncast_dammit() + 1);
+    }
+};
+IteratorPair<ValueIterator<io::FD, IncrFD>> iter_fds();
+
+
+/// open a socket, bind, and listen. Return an fd, or -1 if socket() fails,
+/// but exit if bind() or listen() fails
+Session *make_listen_port(uint16_t port, SessionParsers inferior);
+/// Connect to an address, return a connected socket or -1
+// FIXME - this is IPv4 only!
+Session *make_connection(IP4Address ip, uint16_t port, SessionParsers);
+/// free() the structure and close() the fd
+void delete_session(Session *);
+/// Make a the internal queues bigger
+void realloc_fifo(Session *s, size_t rfifo_size, size_t wfifo_size);
+/// Update all sockets that can be read/written from the queues
+void do_sendrecv(interval_t next);
+/// Call the parser function for every socket that has read data
+void do_parsepacket(void);
+
+template<class T>
+uint8_t *pod_addressof_m(T& structure)
+{
+    static_assert(is_trivially_copyable<T>::value, "Can only byte-copy POD-ish structs");
+    return &reinterpret_cast<uint8_t&>(structure);
+}
+
+template<class T>
+const uint8_t *pod_addressof_c(const T& structure)
+{
+    static_assert(is_trivially_copyable<T>::value, "Can only byte-copy POD-ish structs");
+    return &reinterpret_cast<const uint8_t&>(structure);
+}
+
+
+/// Check how much can be read
+inline
+size_t RFIFOREST(Session *s)
+{
+    return s->rdata_size - s->rdata_pos;
+}
+/// Read from the queue
+inline
+const void *RFIFOP(Session *s, size_t pos)
+{
+    return &s->rdata[s->rdata_pos + pos];
+}
+inline
+uint8_t RFIFOB(Session *s, size_t pos)
+{
+    return *static_cast<const uint8_t *>(RFIFOP(s, pos));
+}
+inline
+uint16_t RFIFOW(Session *s, size_t pos)
+{
+    return *static_cast<const uint16_t *>(RFIFOP(s, pos));
+}
+inline
+uint32_t RFIFOL(Session *s, size_t pos)
+{
+    return *static_cast<const uint32_t *>(RFIFOP(s, pos));
+}
+template<class T>
+void RFIFO_STRUCT(Session *s, size_t pos, T& structure)
+{
+    really_memcpy(pod_addressof_m(structure), static_cast<const uint8_t *>(RFIFOP(s, pos)), sizeof(T));
+}
+inline
+IP4Address RFIFOIP(Session *s, size_t pos)
+{
+    IP4Address o;
+    RFIFO_STRUCT(s, pos, o);
+    return o;
+}
+template<uint8_t len>
+inline
+VString<len-1> RFIFO_STRING(Session *s, size_t pos)
+{
+    const char *const begin = static_cast<const char *>(RFIFOP(s, pos));
+    const char *const end = begin + len-1;
+    const char *const mid = std::find(begin, end, '\0');
+    return XString(begin, mid, nullptr);
+}
+inline
+AString RFIFO_STRING(Session *s, size_t pos, size_t len)
+{
+    const char *const begin = static_cast<const char *>(RFIFOP(s, pos));
+    const char *const end = begin + len;
+    const char *const mid = std::find(begin, end, '\0');
+    return XString(begin, mid, nullptr);
+}
+inline
+void RFIFO_BUF_CLONE(Session *s, uint8_t *buf, size_t len)
+{
+    really_memcpy(buf, static_cast<const uint8_t *>(RFIFOP(s, 0)), len);
+}
+
+/// Done reading
+void RFIFOSKIP(Session *s, size_t len);
+
+/// Read from an arbitrary buffer
+inline
+const void *RBUFP(const uint8_t *p, size_t pos)
+{
+    return p + pos;
+}
+inline
+uint8_t RBUFB(const uint8_t *p, size_t pos)
+{
+    return *static_cast<const uint8_t *>(RBUFP(p, pos));
+}
+inline
+uint16_t RBUFW(const uint8_t *p, size_t pos)
+{
+    return *static_cast<const uint16_t *>(RBUFP(p, pos));
+}
+inline
+uint32_t RBUFL(const uint8_t *p, size_t pos)
+{
+    return *static_cast<const uint32_t *>(RBUFP(p, pos));
+}
+template<class T>
+void RBUF_STRUCT(const uint8_t *p, size_t pos, T& structure)
+{
+    really_memcpy(pod_addressof_m(structure), p + pos, sizeof(T));
+}
+inline
+IP4Address RBUFIP(const uint8_t *p, size_t pos)
+{
+    IP4Address o;
+    RBUF_STRUCT(p, pos, o);
+    return o;
+}
+template<uint8_t len>
+inline
+VString<len-1> RBUF_STRING(const uint8_t *p, size_t pos)
+{
+    const char *const begin = static_cast<const char *>(RBUFP(p, pos));
+    const char *const end = begin + len-1;
+    const char *const mid = std::find(begin, end, '\0');
+    return XString(begin, mid, nullptr);
+}
+inline
+AString RBUF_STRING(const uint8_t *p, size_t pos, size_t len)
+{
+    const char *const begin = static_cast<const char *>(RBUFP(p, pos));
+    const char *const end = begin + len;
+    const char *const mid = std::find(begin, end, '\0');
+    return XString(begin, mid, nullptr);
+}
+
+
+/// Unused - check how much data can be written
+// the existence of this seems scary
+inline
+size_t WFIFOSPACE(Session *s)
+{
+    return s->max_wdata - s->wdata_size;
+}
+/// Write to the queue
+inline
+void *WFIFOP(Session *s, size_t pos)
+{
+    return &s->wdata[s->wdata_size + pos];
+}
+inline
+uint8_t& WFIFOB(Session *s, size_t pos)
+{
+    return *static_cast<uint8_t *>(WFIFOP(s, pos));
+}
+inline
+uint16_t& WFIFOW(Session *s, size_t pos)
+{
+    return *static_cast<uint16_t *>(WFIFOP(s, pos));
+}
+inline
+uint32_t& WFIFOL(Session *s, size_t pos)
+{
+    return *static_cast<uint32_t *>(WFIFOP(s, pos));
+}
+template<class T>
+void WFIFO_STRUCT(Session *s, size_t pos, T& structure)
+{
+    really_memcpy(static_cast<uint8_t *>(WFIFOP(s, pos)), pod_addressof_c(structure), sizeof(T));
+}
+inline
+IP4Address& WFIFOIP(Session *s, size_t pos)
+{
+    static_assert(is_trivially_copyable<IP4Address>::value, "That was the whole point");
+    return *static_cast<IP4Address *>(WFIFOP(s, pos));
+}
+inline
+void WFIFO_STRING(Session *s, size_t pos, XString str, size_t len)
+{
+    char *const begin = static_cast<char *>(WFIFOP(s, pos));
+    char *const end = begin + len;
+    char *const mid = std::copy(str.begin(), str.end(), begin);
+    std::fill(mid, end, '\0');
+}
+inline
+void WFIFO_ZERO(Session *s, size_t pos, size_t len)
+{
+    uint8_t *b = static_cast<uint8_t *>(WFIFOP(s, pos));
+    uint8_t *e = b + len;
+    std::fill(b, e, '\0');
+}
+inline
+void WFIFO_BUF_CLONE(Session *s, const uint8_t *buf, size_t len)
+{
+    really_memcpy(static_cast<uint8_t *>(WFIFOP(s, 0)), buf, len);
+}
+
+/// Finish writing
+void WFIFOSET(Session *s, size_t len);
+
+/// Write to an arbitrary buffer
+inline
+void *WBUFP(uint8_t *p, size_t pos)
+{
+    return p + pos;
+}
+inline
+uint8_t& WBUFB(uint8_t *p, size_t pos)
+{
+    return *static_cast<uint8_t *>(WBUFP(p, pos));
+}
+inline
+uint16_t& WBUFW(uint8_t *p, size_t pos)
+{
+    return *static_cast<uint16_t *>(WBUFP(p, pos));
+}
+inline
+uint32_t& WBUFL(uint8_t *p, size_t pos)
+{
+    return *static_cast<uint32_t *>(WBUFP(p, pos));
+}
+template<class T>
+void WBUF_STRUCT(uint8_t *p, size_t pos, T& structure)
+{
+    really_memcpy(p + pos, pod_addressof_c(structure), sizeof(T));
+}
+inline
+IP4Address& WBUFIP(uint8_t *p, size_t pos)
+{
+    return *static_cast<IP4Address *>(WBUFP(p, pos));
+}
+inline
+void WBUF_STRING(uint8_t *p, size_t pos, XString s, size_t len)
+{
+    char *const begin = static_cast<char *>(WBUFP(p, pos));
+    char *const end = begin + len;
+    char *const mid = std::copy(s.begin(), s.end(), begin);
+    std::fill(mid, end, '\0');
+}
+inline
+void WBUF_ZERO(uint8_t *p, size_t pos, size_t len)
+{
+    uint8_t *b = static_cast<uint8_t *>(WBUFP(p, pos));
+    uint8_t *e = b + len;
+    std::fill(b, e, '\0');
+}
+
+inline
+void RFIFO_WFIFO_CLONE(Session *rs, Session *ws, size_t len)
+{
+    really_memcpy(static_cast<uint8_t *>(WFIFOP(ws, 0)),
+            static_cast<const uint8_t *>(RFIFOP(rs, 0)), len);
+}
+
+#endif // TMWA_NET_SOCKET_HPP
diff --git a/src/net/timer.cpp b/src/net/timer.cpp
new file mode 100644
index 0000000..a85b8be
--- /dev/null
+++ b/src/net/timer.cpp
@@ -0,0 +1,217 @@
+#include "timer.hpp"
+//    timer.cpp - Future event scheduler.
+//
+//    Copyright © ????-2004 Athena Dev Teams
+//    Copyright © 2004-2011 The Mana World Development Team
+//    Copyright © 2011-2014 Ben Longbons <b.r.longbons@gmail.com>
+//
+//    This file is part of The Mana World (Athena server)
+//
+//    This program is free software: you can redistribute it and/or modify
+//    it under the terms of the GNU General Public License as published by
+//    the Free Software Foundation, either version 3 of the License, or
+//    (at your option) any later version.
+//
+//    This program is distributed in the hope that it will be useful,
+//    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.
+//
+//    You should have received a copy of the GNU General Public License
+//    along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+#include <sys/stat.h>
+#include <sys/time.h>
+
+#include <cassert>
+
+#include <algorithm>
+#include <queue>
+
+#include "../strings/zstring.hpp"
+
+#include "../poison.hpp"
+
+struct TimerData
+{
+    /// This will be reset on call, to avoid problems.
+    Timer *owner;
+
+    /// When it will be triggered
+    tick_t tick;
+    /// What will be done
+    timer_func func;
+    /// Repeat rate - 0 for oneshot
+    interval_t interval;
+
+    TimerData(Timer *o, tick_t t, timer_func f, interval_t i)
+    : owner(o)
+    , tick(t)
+    , func(std::move(f))
+    , interval(i)
+    {}
+};
+
+struct TimerCompare
+{
+    /// implement "less than"
+    bool operator() (dumb_ptr<TimerData> l, dumb_ptr<TimerData> r)
+    {
+        // C++ provides a max-heap, but we want
+        // the smallest tick to be the head (a min-heap).
+        return l->tick > r->tick;
+    }
+};
+
+static
+std::priority_queue<dumb_ptr<TimerData>, std::vector<dumb_ptr<TimerData>>, TimerCompare> timer_heap;
+
+
+tick_t gettick_cache;
+
+tick_t milli_clock::now(void) noexcept
+{
+    struct timeval tval;
+    // BUG: This will cause strange behavior if the system clock is changed!
+    // it should be reimplemented in terms of clock_gettime(CLOCK_MONOTONIC, )
+    gettimeofday(&tval, NULL);
+    return gettick_cache = tick_t(std::chrono::seconds(tval.tv_sec)
+            + std::chrono::duration_cast<std::chrono::milliseconds>(
+                std::chrono::microseconds(tval.tv_usec)));
+}
+
+static
+void do_nothing(TimerData *, tick_t)
+{
+}
+
+void Timer::cancel()
+{
+    if (!td)
+        return;
+
+    assert (this == td->owner);
+    td->owner = nullptr;
+    td->func = do_nothing;
+    td->interval = interval_t::zero();
+    td = nullptr;
+}
+
+void Timer::detach()
+{
+    assert (this == td->owner);
+    td->owner = nullptr;
+    td = nullptr;
+}
+
+static
+void push_timer_heap(dumb_ptr<TimerData> td)
+{
+    timer_heap.push(td);
+}
+
+static
+dumb_ptr<TimerData> top_timer_heap(void)
+{
+    if (timer_heap.empty())
+        return dumb_ptr<TimerData>();
+    return timer_heap.top();
+}
+
+static
+void pop_timer_heap(void)
+{
+    timer_heap.pop();
+}
+
+Timer::Timer(tick_t tick, timer_func func, interval_t interval)
+: td(dumb_ptr<TimerData>::make(this, tick, std::move(func), interval))
+{
+    assert (interval >= interval_t::zero());
+
+    push_timer_heap(td);
+}
+
+Timer::Timer(Timer&& t)
+: td(t.td)
+{
+    t.td = nullptr;
+    if (td)
+    {
+        assert (td->owner == &t);
+        td->owner = this;
+    }
+}
+
+Timer& Timer::operator = (Timer&& t)
+{
+    std::swap(td, t.td);
+    if (td)
+    {
+        assert (td->owner == &t);
+        td->owner = this;
+    }
+    if (t.td)
+    {
+        assert (t.td->owner == this);
+        t.td->owner = &t;
+    }
+    return *this;
+}
+
+interval_t do_timer(tick_t tick)
+{
+    /// Number of milliseconds until it calls this again
+    // this says to wait 1 sec if all timers get popped
+    interval_t nextmin = std::chrono::seconds(1);
+
+    while (dumb_ptr<TimerData> td = top_timer_heap())
+    {
+        // while the heap is not empty and
+        if (td->tick > tick)
+        {
+            /// Return the time until the next timer needs to goes off
+            nextmin = td->tick - tick;
+            break;
+        }
+        pop_timer_heap();
+
+        // Prevent destroying the object we're in.
+        // Note: this would be surprising in an interval timer,
+        // but all interval timers do an immediate explicit detach().
+        if (td->owner)
+            td->owner->detach();
+        // If we are too far past the requested tick, call with
+        // the current tick instead to fix reregistration problems
+        if (td->tick + std::chrono::seconds(1) < tick)
+            td->func(td.operator->(), tick);
+        else
+            td->func(td.operator->(), td->tick);
+
+        if (td->interval == interval_t::zero())
+        {
+            td.delete_();
+            continue;
+        }
+        if (td->tick + std::chrono::seconds(1) < tick)
+            td->tick = tick + td->interval;
+        else
+            td->tick += td->interval;
+        push_timer_heap(td);
+    }
+
+    return std::max(nextmin, std::chrono::milliseconds(10));
+}
+
+tick_t file_modified(ZString name)
+{
+    struct stat buf;
+    if (stat(name.c_str(), &buf))
+        return tick_t();
+    return tick_t(std::chrono::seconds(buf.st_mtime));
+}
+
+bool has_timers()
+{
+    return !timer_heap.empty();
+}
diff --git a/src/net/timer.hpp b/src/net/timer.hpp
new file mode 100644
index 0000000..8fd383d
--- /dev/null
+++ b/src/net/timer.hpp
@@ -0,0 +1,50 @@
+#ifndef TMWA_NET_TIMER_HPP
+#define TMWA_NET_TIMER_HPP
+//    timer.hpp - Future event scheduler.
+//
+//    Copyright © ????-2004 Athena Dev Teams
+//    Copyright © 2004-2011 The Mana World Development Team
+//    Copyright © 2011-2014 Ben Longbons <b.r.longbons@gmail.com>
+//
+//    This file is part of The Mana World (Athena server)
+//
+//    This program is free software: you can redistribute it and/or modify
+//    it under the terms of the GNU General Public License as published by
+//    the Free Software Foundation, either version 3 of the License, or
+//    (at your option) any later version.
+//
+//    This program is distributed in the hope that it will be useful,
+//    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.
+//
+//    You should have received a copy of the GNU General Public License
+//    along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+# include "timer.t.hpp"
+
+# include "fwd.hpp"
+
+# include "../strings/fwd.hpp"
+
+// updated automatically when using milli_clock::now()
+// which is done only by core.cpp
+extern tick_t gettick_cache;
+
+inline
+tick_t gettick(void)
+{
+    return gettick_cache;
+}
+
+/// Do all timers scheduled before tick, and return the number of
+/// milliseconds until the next timer happens
+interval_t do_timer(tick_t tick);
+
+/// Stat a file, and return its modification time, truncated to seconds.
+tick_t file_modified(ZString name);
+
+/// Check if there are any events at all scheduled.
+bool has_timers();
+
+#endif // TMWA_NET_TIMER_HPP
diff --git a/src/net/timer.t.hpp b/src/net/timer.t.hpp
new file mode 100644
index 0000000..91d4b74
--- /dev/null
+++ b/src/net/timer.t.hpp
@@ -0,0 +1,88 @@
+#ifndef TMWA_NET_TIMER_T_HPP
+#define TMWA_NET_TIMER_T_HPP
+//    timer.t.hpp - Future event scheduler.
+//
+//    Copyright © ????-2004 Athena Dev Teams
+//    Copyright © 2004-2011 The Mana World Development Team
+//    Copyright © 2011-2014 Ben Longbons <b.r.longbons@gmail.com>
+//
+//    This file is part of The Mana World (Athena server)
+//
+//    This program is free software: you can redistribute it and/or modify
+//    it under the terms of the GNU General Public License as published by
+//    the Free Software Foundation, either version 3 of the License, or
+//    (at your option) any later version.
+//
+//    This program is distributed in the hope that it will be useful,
+//    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.
+//
+//    You should have received a copy of the GNU General Public License
+//    along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+# include "fwd.hpp"
+
+# include <chrono>
+# include <functional>
+
+# include "../generic/dumb_ptr.hpp"
+
+/// An implementation of the C++ "clock" concept, exposing
+/// durations in milliseconds.
+class milli_clock
+{
+public:
+    typedef std::chrono::milliseconds duration;
+    typedef duration::rep rep;
+    typedef duration::period period;
+    typedef std::chrono::time_point<milli_clock, duration> time_point;
+    static const bool is_steady = true; // assumed - not necessarily true
+
+    static time_point now() noexcept;
+};
+
+/// A point in time.
+typedef milli_clock::time_point tick_t;
+/// The difference between two points in time.
+typedef milli_clock::duration interval_t;
+/// (to get additional arguments, use std::bind or a lambda).
+typedef std::function<void (TimerData *, tick_t)> timer_func;
+
+class Timer
+{
+    friend struct TimerData;
+    dumb_ptr<TimerData> td;
+
+    Timer(const Timer&) = delete;
+    Timer& operator = (const Timer&) = delete;
+public:
+    /// Don't own anything yet.
+    Timer() = default;
+    /// Schedule a timer for the given tick.
+    /// If you do not wish to keep track of it, call disconnect().
+    /// Otherwise, you may cancel() or replace (operator =) it later.
+    ///
+    /// If the interval argument is given, the timer will reschedule
+    /// itself again forever. Otherwise, it will disconnect() itself
+    /// just BEFORE it is called.
+    Timer(tick_t tick, timer_func func, interval_t interval=interval_t::zero());
+
+    Timer(Timer&& t);
+    Timer& operator = (Timer&& t);
+    ~Timer() { cancel(); }
+
+    /// Cancel the delivery of this timer's function, and make it falsy.
+    /// Implementation note: this doesn't actually remove it, just sets
+    /// the functor to do_nothing, and waits for the tick before removing.
+    void cancel();
+    /// Make it falsy without cancelling the timer,
+    void detach();
+
+    /// Check if there is a timer connected.
+    explicit operator bool() { return bool(td); }
+    /// Check if there is no connected timer.
+    bool operator !() { return !td; }
+};
+
+#endif // TMWA_NET_TIMER_T_HPP