diff --git a/Makefile b/Makefile
index fb4a84ff..861ac91f 100644
--- a/Makefile
+++ b/Makefile
@@ -1,11 +1,11 @@
 test/echo-server: test/echo-server.c ol.a
-	$(CC) -o test/echo-server test/echo-server.c ol.a -lm
+	$(CC) -g -o test/echo-server test/echo-server.c ol.a -lm
 
 ol.a: ol-unix.o ev/ev.o
 	$(AR) rcs ol.a ol-unix.o ev/ev.o
 
 ol-unix.o: ol-unix.c ol.h ol-unix.h
-	$(CC) -c ol-unix.c -o ol-unix.o -lm
+	$(CC) -g -c ol-unix.c -o ol-unix.o -lm
 
 ev/ev.o: ev/config.h ev/ev.c
 	$(MAKE) -C ev
diff --git a/ol-unix.c b/ol-unix.c
index 2b7b769a..c70ffc9d 100644
--- a/ol-unix.c
+++ b/ol-unix.c
@@ -11,10 +11,15 @@ void ol_tcp_connect(ol_handle* handle, ol_req* req);
 int ol_close_error(ol_handle* handle, ol_err err);
 
 
-static int ol_err_new(int e) {
+static ol_err ol_err_new(ol_handle* handle, int e) {
+  handle->_.err = e;
   return e;
 }
 
+ol_err ol_err_last(ol_handle *handle) {
+  return handle->_.err;
+}
+
 
 struct sockaddr_in ol_ip4_addr(char *ip, int port) {
   struct sockaddr_in addr;
@@ -32,8 +37,13 @@ int ol_close(ol_handle* handle) {
 }
 
 
+void ol_init() {
+  ev_default_loop(0);
+}
+
+
 int ol_run() {
-  ev_run(0);
+  ev_run(EV_DEFAULT_ 0);
 }
 
 
@@ -42,6 +52,8 @@ ol_handle* ol_handle_new(ol_close_cb close_cb, void* data) {
   handle->close_cb = close_cb;
   handle->data = data;
 
+  handle->_.fd = -1;
+
   ev_init(&handle->_.read_watcher, ol_tcp_io);
   ev_init(&handle->_.write_watcher, ol_tcp_io);
 
@@ -49,33 +61,114 @@ ol_handle* ol_handle_new(ol_close_cb close_cb, void* data) {
 }
 
 
+int ol_tcp_lazy_open(ol_handle* handle, int domain) {
+  assert(handle->_.fd < 0);
+
+  /* Lazily allocate a file descriptor for this handle */
+  handle->_.fd = socket(domain, SOCK_STREAM, 0);
+  int yes = 1;
+  setsockopt(handle->_.fd, SOL_SOCKET, SO_REUSEADDR, &yes, sizeof(int));
+  fcntl(handle->_.fd, F_SETFL, O_NONBLOCK);
+
+  return 0;
+}
+
+
 int ol_bind(ol_handle* handle, struct sockaddr* addr) {
   int addrsize;
+  int domain;
 
   if (addr->sa_family == AF_INET) {
     addrsize = sizeof(struct sockaddr_in);
+    domain = AF_INET;
   } else if (addr->sa_family == AF_INET6) {
     addrsize = sizeof(struct sockaddr_in6);
+    domain = AF_INET6;
   } else {
     assert(0);
     return -1;
   }
 
-  int r = bind(handle->_.fd, addr, addrsize);
+  int r = 0;
 
-  return ol_err_new(r);
+  if (handle->_.fd < 0) {
+    r = ol_tcp_lazy_open(handle, domain);
+    if (r) {
+      return ol_err_new(handle, r);
+    }
+  }
+
+  r = bind(handle->_.fd, addr, addrsize);
+
+  return ol_err_new(handle, r);
+}
+
+
+ol_handle* ol_tcp_open(ol_handle* parent, int fd) {
+  ol_handle* h = ol_handle_new(NULL, NULL);
+  h->_.fd = fd;
+  return h;
+}
+
+
+void ol_server_io(EV_P_ ev_io* watcher, int revents) {
+  ol_handle* handle = watcher->data;
+
+  assert(revents == EV_READ);
+
+  while (1) {
+    struct sockaddr addr;
+    socklen_t addrlen;
+    int fd = accept(handle->_.fd, &addr, &addrlen);
+
+    if (fd < 0) {
+      if (errno == EAGAIN) {
+        return; // No problem.
+      } else if (errno == EMFILE) {
+        // TODO special trick. unlock reserved socket, accept, close.
+        return;
+      } else {
+        ol_close_error(handle, ol_err_new(handle, errno));
+      }
+    } else {
+      if (handle->accept_cb) {
+        ol_handle* new_client = ol_tcp_open(handle, fd);
+        if (!new_client) {
+          ol_close_error(handle, ol_err_last(handle));
+          return;
+        }
+
+        handle->accept_cb(handle, new_client);
+      } else {
+        close(fd);
+      }
+    }
+  }
 }
 
 
 int ol_listen(ol_handle* handle, int backlog, ol_accept_cb cb) {
+  if (handle->_.fd < 0) {
+    /* Lazily allocate a file descriptor for this handle */
+    handle->_.fd = socket(AF_INET, SOCK_STREAM, 0);
+  }
+
   int r = listen(handle->_.fd, backlog);
+  if (r < 0) {
+    return ol_err_new(handle, errno);
+  }
+
   handle->accept_cb = cb;
-  return ol_err_new(r);
+  ev_io_init(&handle->_.read_watcher, ol_server_io, handle->_.fd, EV_READ);
+  ev_io_start(EV_DEFAULT_ &handle->_.read_watcher);
+  handle->_.read_watcher.data = handle;
+
+  return 0;
 }
 
 
 int ol_close_error(ol_handle* handle, ol_err err) {
-  ev_io_stop(&handle->_.read_watcher);
+  ev_io_stop(EV_DEFAULT_ &handle->_.read_watcher);
   close(handle->_.fd);
   handle->_.fd = -1;
 
@@ -121,9 +214,9 @@ void ol_tcp_connect(ol_handle* handle, ol_req* req) {
     ol_connect_cb connect_cb = req->connect_cb;
     if (connect_cb) {
       if (req->_.local) {
-        connect_cb(NULL, ol_err_new(0));
+        connect_cb(NULL, ol_err_new(handle, 0));
       } else {
-        connect_cb(req, ol_err_new(0));
+        connect_cb(req, ol_err_new(handle, 0));
       }
     }
 
@@ -139,11 +232,13 @@ void ol_tcp_connect(ol_handle* handle, ol_req* req) {
     return;
 
   } else {
+    ol_err err = ol_err_new(handle, error);
+
     if (req->_.connect_cb) {
-      req->_.connect_cb(req, ol_err_new(error));
+      req->_.connect_cb(req, err);
     }
 
-    ol_close_error(handle, ol_err_new(error));
+    ol_close_error(handle, err);
   }
 }
 
@@ -165,16 +260,16 @@ ol_req* ol_req_maybe_alloc(ol_handle* handle, ol_req* in_req) {
 
 int ol_connect(ol_handle* handle, ol_req *req_in, struct sockaddr* addr) {
   if (handle->_.connect_req) {
-    return ol_err_new(EALREADY);
+    return ol_err_new(handle, EALREADY);
   }
 
   if (handle->type != OL_TCP) {
-    return ol_err_new(ENOTSOCK);
+    return ol_err_new(handle, ENOTSOCK);
   }
 
   ol_req *req = ol_req_maybe_alloc(handle, req_in);
   if (!req) {
-    return ol_err_new(ENOMEM);
+    return ol_err_new(handle, ENOMEM);
   }
 
   handle->_.connect_req = req;
@@ -195,16 +290,16 @@ int ol_connect(ol_handle* handle, ol_req *req_in, struct sockaddr* addr) {
 
   /* socket(2) failed */
   if (handle->_.fd < 0) {
-    return ol_err_new(errno);
+    return ol_err_new(handle, errno);
   }
 
   int r = connect(handle->_.fd, addr, addrsize);
 
   ev_io_init(&handle->_.read_watcher, ol_tcp_io, handle->_.fd, EV_READ);
   ev_io_init(&handle->_.write_watcher, ol_tcp_io, handle->_.fd, EV_WRITE);
-  ev_io_start(&handle->_.read_watcher);
+  ev_io_start(EV_DEFAULT_ &handle->_.read_watcher);
 
-  return ol_err_new(r);
+  return ol_err_new(handle, r);
 }
 
 int ol_write(ol_handle* handle, ol_req *req, ol_buf* bufs, int bufcnt) {
@@ -223,9 +318,60 @@ int ol_write2(ol_handle* handle, const char* msg) {
 }
 
 
-int ol_read(ol_handle* handle, ol_req *req, ol_buf* bufs, int bufcnt) {
-  // stub
-  assert(0);
+void ol_req_append(ol_handle* handle, ol_req *req) {
+  ngx_queue_insert_tail(&handle->read_reqs, &req->read_queue);
+}
+
+
+int ol_read(ol_handle* handle, ol_req *req_in, ol_buf* bufs, int bufcnt) {
+  assert(handle->_.fd >= 0);
+
+  if (!ngx_queue_empty(&handle->read_reqs)) {
+    /* There are already pending read_reqs. We must get in line. */
+    assert(ev_is_active(&handle->read_watcher));
+
+    ol_req *req = ol_req_maybe_alloc(handle, req_in);
+    if (!req) {
+      return ol_err_new(handle, ENOMEM);
+    }
+
+    ol_req_append(handle, req);
+
+    return ol_err_new(handle, EINPROGRESS);
+
+  } else {
+    /* Attempt to read immediately */
+    ssize_t nread = readv(handle->_.fd, (struct iovec) bufs, bufcnt);
+
+    if (nread < 0) {
+
+    }
+
+    ssize_t buftotal;
+    for (int i = 0; i < bufcnt; i++) {
+      buftotal += bufs.len;
+      if (buftotal == nread) {
+        /* We read everything */
+        if (req_in.read_cb) {
+          req_in.read_cb(handle, req_in, nread, 0);
+        }
+      }
+
+      if (buftotal > nread) break;
+    }
+
+
+    ol_req *req = ol_req_maybe_alloc(handle, req_in);
+    if (!req) {
+      return ol_err_new(handle, ENOMEM);
+    }
+
+    // blah
+  }
+
+
+  ngx_queue_insert_tail();
+
   return 0;
 }
 
diff --git a/ol-unix.h b/ol-unix.h
index 8750b0ae..39bad08a 100644
--- a/ol-unix.h
+++ b/ol-unix.h
@@ -3,7 +3,6 @@
 
 #include "ngx-queue.h"
 
-#define EV_MULTIPLICITY 0
 #include "ev/ev.h"
 
 #include <sys/types.h>
@@ -42,6 +41,8 @@ typedef struct {
 typedef struct {
   int fd;
 
+  ol_err err;
+
   ol_read_cb read_cb;
   ol_close_cb close_cb;
 
diff --git a/ol.h b/ol.h
index b7e32db7..53748b80 100644
--- a/ol.h
+++ b/ol.h
@@ -73,6 +73,7 @@ struct ol_req_s {
 };
 
 
+void ol_init();
 int ol_run();
 
 ol_handle* ol_handle_new(ol_close_cb close_cb, void* data);
diff --git a/test/echo-server.c b/test/echo-server.c
index 96227a44..9adf70ff 100644
--- a/test/echo-server.c
+++ b/test/echo-server.c
@@ -80,11 +80,23 @@ void on_accept(ol_handle* server, ol_handle* new_client) {
 
 
 int main(int argc, char** argv) {
+  ol_init();
+
   ol_handle* server = ol_handle_new(on_close, NULL);
 
   struct sockaddr_in addr = ol_ip4_addr("0.0.0.0", 8000);
-  ol_bind(server, (struct sockaddr*) &addr);
-  ol_listen(server, 128, on_accept);
+
+  int r = ol_bind(server, (struct sockaddr*) &addr);
+  if (r) {
+    fprintf(stderr, "Bind error\n");
+    return 1;
+  }
+
+  r = ol_listen(server, 128, on_accept);
+  if (r) {
+    fprintf(stderr, "Listen error\n");
+    return 1;
+  }
 
   ol_run();