diff --git a/TODO b/TODO
index 8f8486ae..fe9b015b 100644
--- a/TODO
+++ b/TODO
@@ -27,18 +27,6 @@ Core wayland protocol
       send a NULL type or x-wayland/root-something type if the source
       offers that.
 
-   Races between pointer motion, ending the drag, the target sending
-   accept request and the source receiving the target event.
-
-    - We've sent a drag focus or motion event to the source, but
-      haven't received an accept request corresponding to that event
-      and now the button is release.  The compositor could wait for
-      the source to reply to outstanding focus/motion events before
-      sending the finish event to the source.  Or we could send the
-      finish event through the source so that it needs to reply to the
-      finish event too.  Either way, the state of the drag blocks on
-      the client.  What if we drag to a client that doesn't doo dnd?
-
    How do we animate the drag icon back to the drag origin in case of
    a failed drag?
 
@@ -48,18 +36,6 @@ Core wayland protocol
    that will participate in dnd.  Or just assume client is not
    participating until we receive an accept request.
 
-   May need to look at all offer events before we can decide which one
-   to go with.  Problem is, we don't know when we've seen that last
-   offer event.
-
-   Create transient dnd object when a client starts a drag.  Announce
-   the dnd object to clients first time the drag enters one of its
-   surfaces.  Track if we've already announced the object by comparing
-   the drag start timestamp/serial with the clients last-event
-   timestamp/serial?  Wont work if we send other events to a client
-   after creating the drag object.  Maybe just keep the transient
-   object on the initiator side?
-
  - Pointer image issue:
 
     - A touch input device doesn't have a pointer; indicate that
diff --git a/clients/dnd.c b/clients/dnd.c
index a53caa28..df4d8c57 100644
--- a/clients/dnd.c
+++ b/clients/dnd.c
@@ -44,12 +44,21 @@ struct dnd {
 	struct display *display;
 	uint32_t key;
 	struct item *items[16];
+};
 
-	struct wl_buffer *translucent_buffer;
-	struct wl_buffer *opaque_buffer;
+struct dnd_drag {
+	cairo_surface_t *translucent;
+	cairo_surface_t *opaque;
 	int hotspot_x, hotspot_y;
-	uint32_t tag;
+	struct dnd *dnd;
+	struct input *input;
+};
+
+struct dnd_offer {
+	struct dnd *dnd;
+	struct wl_array types;
 	const char *drag_type;
+	uint32_t tag;
 };
 
 struct item {
@@ -108,7 +117,7 @@ item_create(struct display *display, int x, int y)
 		cairo_curve_to(cr,
 			       x1 - y1 * u, y1 + x1 * u,
 			       x2 + y2 * v, y2 - x2 * v,
-			       x2, y2);			       
+			       x2, y2);
 
 		cairo_curve_to(cr,
 			       x2 - y2 * v, y2 + x2 * v,
@@ -219,90 +228,130 @@ dnd_get_item(struct dnd *dnd, int32_t x, int32_t y)
 }
 
 static void
-drag_handle_device(void *data,
-		   struct wl_drag *drag, struct wl_input_device *device)
+drag_target(void *data,
+	    struct wl_drag *drag, const char *mime_type)
 {
+	struct dnd_drag *dnd_drag = data;
+	struct dnd *dnd = dnd_drag->dnd;
+	struct wl_input_device *device;
+	cairo_surface_t *surface;
+	struct wl_buffer *buffer;
+
+	fprintf(stderr, "target %s\n", mime_type);
+	device = input_get_input_device(dnd_drag->input);
+	if (mime_type)
+		surface = dnd_drag->opaque;
+	else
+		surface = dnd_drag->translucent;
+
+	buffer = display_get_buffer_for_surface(dnd->display, surface);
+	wl_input_device_attach(device, buffer,
+			       dnd_drag->hotspot_x, dnd_drag->hotspot_y);
 }
 
 static void
-drag_pointer_focus(void *data,
-		   struct wl_drag *drag,
-		   uint32_t time, struct wl_surface *surface,
-		   int32_t x, int32_t y, int32_t surface_x, int32_t surface_y)
+drag_finish(void *data, struct wl_drag *drag, int fd)
 {
-	struct dnd *dnd = data;
-
-	/* FIXME: We need the offered types before we get the
-	 * pointer_focus event so we know which one we want and can
-	 * send the accept request back. */
+	struct dnd_drag *dnd_drag = data;
+	char text[] = "[drop data]";
 
-	fprintf(stderr, "drag pointer focus %p\n", surface);
+	fprintf(stderr, "got 'finish', fd %d, sending message\n", fd);
 
-	if (!surface) {
-		dnd->drag_type = NULL;
-		return;
-	}
+	write(fd, text, sizeof text);
+	close(fd);
 
-	if (!dnd_get_item(dnd, surface_x, surface_y)) {
-		wl_drag_accept(drag, "text/plain");
-		dnd->drag_type = "text/plain";
-	} else {
-		wl_drag_accept(drag, NULL);
-		dnd->drag_type = NULL;
-	}
+	/* The 'finish' event marks the end of the session on the drag
+	 * source side and we need to clean up the drag object created
+	 * and the local state. */
+	wl_drag_destroy(drag);
+	cairo_surface_destroy(dnd_drag->translucent);
+	cairo_surface_destroy(dnd_drag->opaque);
+	free(dnd_drag);
 }
 
+static const struct wl_drag_listener drag_listener = {
+	drag_target,
+	drag_finish
+};
+
 static void
-drag_offer(void *data,
-	   struct wl_drag *drag, const char *type)
+drag_offer_offer(void *data,
+		 struct wl_drag_offer *offer, const char *type)
 {
-	fprintf(stderr, "drag offer %s\n", type);
+	struct dnd_offer *dnd_offer = data;
+	char **p;
+
+	p = wl_array_add(&dnd_offer->types, sizeof *p);
+	if (p)
+		*p = strdup(type);
 }
 
 static void
-drag_motion(void *data,
-	    struct wl_drag *drag,
-	    uint32_t time,
-	    int32_t x, int32_t y, int32_t surface_x, int32_t surface_y)
+drag_offer_pointer_focus(void *data,
+			 struct wl_drag_offer *offer,
+			 uint32_t time, struct wl_surface *surface,
+			 int32_t x, int32_t y,
+			 int32_t surface_x, int32_t surface_y)
 {
-	struct dnd *dnd = data;
+	struct dnd_offer *dnd_offer = data;
+	struct window *window;
+	char **p, **end;
 
-	/* FIXME: Need to correlate this with the offer event.
-	 * Problem is, we don't know when we've seen that last offer
-	 * event, and we might need to look at all of them before we
-	 * can decide which one to go with. */
-	if (!dnd_get_item(dnd, surface_x, surface_y)) {
-		wl_drag_accept(drag, "text/plain");
-		dnd->drag_type = "text/plain";
+	/* The last event in a dnd session is pointer_focus with a
+	 * NULL surface, whether or not we get the drop event.  We
+	 * need to clean up the dnd_offer proxy and whatever state we
+	 * allocated. */
+	if (!surface) {
+		fprintf(stderr, "pointer focus NULL, session over\n");
+		wl_array_release(&dnd_offer->types);
+		free(dnd_offer);
+		wl_drag_offer_destroy(offer);
+		return;
+	}
+
+	fprintf(stderr, "drag pointer focus %p\n", surface);
+	fprintf(stderr, "offered types:\n");
+	end = dnd_offer->types.data + dnd_offer->types.size;
+	for (p = dnd_offer->types.data; p < end; p++)
+		fprintf(stderr, "\%s\n", *p);
+
+	window = wl_surface_get_user_data(surface);
+	dnd_offer->dnd = window_get_user_data(window);
+
+	if (!dnd_get_item(dnd_offer->dnd, surface_x, surface_y)) {
+		wl_drag_offer_accept(offer, time, "text/plain");
+		dnd_offer->drag_type = "text/plain";
 	} else {
-		wl_drag_accept(drag, NULL);
-		dnd->drag_type = NULL;
+		wl_drag_offer_accept(offer, time, NULL);
+		dnd_offer->drag_type = NULL;
 	}
 }
 
 static void
-drag_target(void *data,
-	    struct wl_drag *drag, const char *mime_type)
+drag_offer_motion(void *data,
+		  struct wl_drag_offer *offer, uint32_t time,
+		  int32_t x, int32_t y, int32_t surface_x, int32_t surface_y)
 {
-	struct dnd *dnd = data;
-	struct input *input;
-	struct wl_input_device *device;
+	struct dnd_offer *dnd_offer = data;
+	struct dnd *dnd = dnd_offer->dnd;
 
-	fprintf(stderr, "target %s\n", mime_type);
-	input = wl_drag_get_user_data(drag);
-	device = input_get_input_device(input);
-	if (mime_type)
-		wl_input_device_attach(device, dnd->opaque_buffer,
-				       dnd->hotspot_x, dnd->hotspot_y);
-	else
-		wl_input_device_attach(device, dnd->translucent_buffer,
-				       dnd->hotspot_x, dnd->hotspot_y);
+	if (!dnd_get_item(dnd, surface_x, surface_y)) {
+		fprintf(stderr, "drag offer motion %d, %d, accepting\n",
+			surface_x, surface_y);
+		wl_drag_offer_accept(offer, time, "text/plain");
+		dnd_offer->drag_type = "text/plain";
+	} else {
+		fprintf(stderr, "drag offer motion %d, %d, declining\n",
+			surface_x, surface_y);
+		wl_drag_offer_accept(offer, time, NULL);
+		dnd_offer->drag_type = NULL;
+	}
 }
 
 static gboolean
 drop_io_func(GIOChannel *source, GIOCondition condition, gpointer data)
 {
-	struct dnd *dnd = data;
+	struct dnd_offer *dnd_offer = data;
 	char buffer[256];
 	int fd;
 	unsigned int len;
@@ -312,7 +361,7 @@ drop_io_func(GIOChannel *source, GIOCondition condition, gpointer data)
 	fprintf(stderr, "read %d bytes: %s\n", len, buffer);
 	fd = g_io_channel_unix_get_fd(source);
 	close(fd);
-	g_source_remove(dnd->tag);
+	g_source_remove(dnd_offer->tag);
 
 	g_io_channel_unref(source);
 
@@ -320,13 +369,13 @@ drop_io_func(GIOChannel *source, GIOCondition condition, gpointer data)
 }
 
 static void
-drag_drop(void *data, struct wl_drag *drag)
+drag_offer_drop(void *data, struct wl_drag_offer *offer)
 {
-	struct dnd *dnd = data;
-	int p[2];
+	struct dnd_offer *dnd_offer = data;
 	GIOChannel *channel;
+	int p[2];
 
-	if (!dnd->drag_type) {
+	if (!dnd_offer->drag_type) {
 		fprintf(stderr, "got 'drop', but no target\n");
 		/* FIXME: Should send response so compositor and
 		 * source knows it's over. Can't send -1 to indicate
@@ -338,38 +387,39 @@ drag_drop(void *data, struct wl_drag *drag)
 	fprintf(stderr, "got 'drop', sending write end of pipe\n");
 
 	pipe(p);
-	wl_drag_receive(drag, p[1]);
+	wl_drag_offer_receive(offer, p[1]);
 	close(p[1]);
 
 	channel = g_io_channel_unix_new(p[0]);
-	dnd->tag = g_io_add_watch(channel, G_IO_IN, drop_io_func, dnd);
+	dnd_offer->tag = g_io_add_watch(channel, G_IO_IN,
+					drop_io_func, dnd_offer);
 }
 
+static const struct wl_drag_offer_listener drag_offer_listener = {
+	drag_offer_offer,
+	drag_offer_pointer_focus,
+	drag_offer_motion,
+	drag_offer_drop,
+};
+
 static void
-drag_finish(void *data, struct wl_drag *drag, int fd)
+drag_offer_handler(struct wl_drag_offer *offer, struct display *display)
 {
-	char text[] = "[drop data]";
+	struct dnd_offer *dnd_offer;
 
-	fprintf(stderr, "got 'finish', fd %d, sending message\n", fd);
+	dnd_offer = malloc(sizeof *dnd_offer);
+	if (dnd_offer == NULL)
+		return;
 
-	write(fd, text, sizeof text);
-	close(fd);
+	wl_drag_offer_add_listener(offer, &drag_offer_listener, dnd_offer);
+	wl_array_init(&dnd_offer->types);
 }
 
-static const struct wl_drag_listener drag_listener = {
-	drag_handle_device,
-	drag_pointer_focus,
-	drag_offer,
-	drag_motion,
-	drag_target,
-	drag_drop,
-	drag_finish
-};
-
 static cairo_surface_t *
-create_drag_cursor(struct dnd *dnd, struct item *item, int32_t x, int32_t y,
-		   double opacity)
+create_drag_cursor(struct dnd_drag *dnd_drag,
+		   struct item *item, int32_t x, int32_t y, double opacity)
 {
+	struct dnd *dnd = dnd_drag->dnd;
 	cairo_surface_t *surface, *pointer;
 	int32_t pointer_width, pointer_height, hotspot_x, hotspot_y;
 	struct rectangle rectangle;
@@ -409,8 +459,8 @@ create_drag_cursor(struct dnd *dnd, struct item *item, int32_t x, int32_t y,
 	display_flush_cairo_device(dnd->display);
 	cairo_destroy(cr);
 
-	dnd->hotspot_x = pointer_width + x - item->x;
-	dnd->hotspot_y = pointer_height + y - item->y;
+	dnd_drag->hotspot_x = pointer_width + x - item->x;
+	dnd_drag->hotspot_y = pointer_height + y - item->y;
 
 	return surface;
 }
@@ -423,8 +473,8 @@ dnd_button_handler(struct window *window,
 	struct dnd *dnd = data;
 	int32_t x, y;
 	struct item *item;
-	cairo_surface_t *surface;
 	struct rectangle rectangle;
+	struct dnd_drag *dnd_drag;
 
 	window_get_child_rectangle(dnd->window, &rectangle);
 	input_get_position(input, &x, &y);
@@ -435,18 +485,17 @@ dnd_button_handler(struct window *window,
 	if (item && state == 1) {
 		fprintf(stderr, "start drag, item %p\n", item);
 
-		surface = create_drag_cursor(dnd, item, x, y, 1);
-		dnd->opaque_buffer =
-			display_get_buffer_for_surface(dnd->display, surface);
-
-		surface = create_drag_cursor(dnd, item, x, y, 0.2);
-		dnd->translucent_buffer =
-			display_get_buffer_for_surface(dnd->display, surface);
+		dnd_drag = malloc(sizeof *dnd_drag);
+		dnd_drag->dnd = dnd;
+		dnd_drag->input = input;
 
-		window_start_drag(window, input, time);
+		dnd_drag->opaque =
+			create_drag_cursor(dnd_drag, item, x, y, 1);
+		dnd_drag->translucent =
+			create_drag_cursor(dnd_drag, item, x, y, 0.2);
 
-		/* FIXME: We leak the surface because we can't free it
-		 * until the server has referenced it. */
+		window_start_drag(window, input, time,
+				  &drag_listener, dnd_drag);
 	}
 }
 
@@ -509,8 +558,6 @@ dnd_create(struct display *display)
 	rectangle.height = 4 * (item_height + item_padding) + item_padding;
 	window_set_child_size(dnd->window, &rectangle);
 
-	display_add_drag_listener(display, &drag_listener, dnd);
-
 	dnd_draw(dnd);
 
 	return dnd;
@@ -532,6 +579,8 @@ main(int argc, char *argv[])
 
 	d = display_create(&argc, &argv, option_entries);
 
+	display_set_drag_offer_handler(d, drag_offer_handler);
+
 	dnd = dnd_create (d);
 
 	display_run(d);
diff --git a/clients/window.c b/clients/window.c
index b6074e57..5a2f2475 100644
--- a/clients/window.c
+++ b/clients/window.c
@@ -71,6 +71,8 @@ struct display {
 	cairo_surface_t *active_frame, *inactive_frame, *shadow;
 	struct xkb_desc *xkb;
 	cairo_surface_t **pointer_surfaces;
+
+	display_drag_offer_handler_t drag_offer_handler;
 };
 
 struct window {
@@ -107,7 +109,6 @@ struct window {
 struct input {
 	struct display *display;
 	struct wl_input_device *input_device;
-	struct wl_drag *drag;
 	struct window *pointer_focus;
 	struct window *keyboard_focus;
 	uint32_t modifiers;
@@ -744,25 +745,21 @@ input_get_input_device(struct input *input)
 	return input->input_device;
 }
 
-void
-display_add_drag_listener(struct display *display,
-			  const struct wl_drag_listener *drag_listener,
-			  void *data)
+struct wl_drag *
+window_start_drag(struct window *window, struct input *input, uint32_t time,
+		  const struct wl_drag_listener *listener, void *data)
 {
-	struct input *input;
-
-	wl_list_for_each(input, &display->input_list, link)
-		wl_drag_add_listener(input->drag, drag_listener, data);
-}
+	struct wl_drag *drag;
 
-void
-window_start_drag(struct window *window, struct input *input, uint32_t time)
-{
 	cairo_device_flush (window->display->device);
 
-	wl_drag_prepare(input->drag, window->surface, time);
-	wl_drag_offer(input->drag, "text/plain");
-	wl_drag_activate(input->drag);
+	drag = wl_shell_create_drag(window->display->shell);
+	wl_drag_offer(drag, "text/plain");
+	wl_drag_offer(drag, "text/html");
+	wl_drag_activate(drag, window->surface, input->input_device, time);
+	wl_drag_add_listener(drag, listener, data);
+
+	return drag;
 }
 
 static void
@@ -902,6 +899,12 @@ window_set_user_data(struct window *window, void *data)
 	window->user_data = data;
 }
 
+void *
+window_get_user_data(struct window *window)
+{
+	return window->user_data;
+}
+
 void
 window_set_resize_handler(struct window *window,
 			  window_resize_handler_t handler)
@@ -1099,71 +1102,12 @@ display_add_input(struct display *d, uint32_t id)
 	wl_input_device_set_user_data(input->input_device, input);
 }
 
-static void
-drag_handle_device(void *data,
-		   struct wl_drag *drag, struct wl_input_device *device)
-{
-	struct input *input;
-
-	input = wl_input_device_get_user_data(device);
-	input->drag = drag;
-	wl_drag_set_user_data(drag, input);
-}
-
-static void
-drag_pointer_focus(void *data,
-		   struct wl_drag *drag,
-		   uint32_t time, struct wl_surface *surface,
-		   int32_t x, int32_t y, int32_t surface_x, int32_t surface_y)
-{
-}
-
-static void
-drag_offer(void *data,
-	   struct wl_drag *drag, const char *type)
-{
-}
-
-static void
-drag_motion(void *data,
-	    struct wl_drag *drag,
-	    uint32_t time,
-	    int32_t x, int32_t y, int32_t surface_x, int32_t surface_y)
-{
-}
-
-static void
-drag_target(void *data,
-	    struct wl_drag *drag, const char *mime_type)
-{
-}
-
-static void
-drag_drop(void *data, struct wl_drag *drag)
-{
-}
-
-static void
-drag_finish(void *data, struct wl_drag *drag, int fd)
-{
-}
-
-static const struct wl_drag_listener drag_listener = {
-	drag_handle_device,
-	drag_pointer_focus,
-	drag_offer,
-	drag_motion,
-	drag_target,
-	drag_drop,
-	drag_finish
-};
-
 static void
 display_handle_global(struct wl_display *display, uint32_t id,
 		      const char *interface, uint32_t version, void *data)
 {
 	struct display *d = data;
-	struct wl_drag *drag;
+	struct wl_drag_offer *offer;
 
 	if (strcmp(interface, "compositor") == 0) {
 		d->compositor = wl_compositor_create(display, id);
@@ -1180,9 +1124,9 @@ display_handle_global(struct wl_display *display, uint32_t id,
 	} else if (strcmp(interface, "drm") == 0) {
 		d->drm = wl_drm_create(display, id);
 		wl_drm_add_listener(d->drm, &drm_listener, d);
-	} else if (strcmp(interface, "drag") == 0) {
-		drag = wl_drag_create(display, id);
-		wl_drag_add_listener(drag, &drag_listener, NULL);
+	} else if (strcmp(interface, "drag_offer") == 0) {
+		offer = wl_drag_offer_create(display, id);
+		d->drag_offer_handler(offer, d);
 	}
 }
 
@@ -1363,3 +1307,10 @@ display_run(struct display *d)
 {
 	g_main_loop_run(d->loop);
 }
+
+void
+display_set_drag_offer_handler(struct display *display,
+			       display_drag_offer_handler_t handler)
+{
+	display->drag_offer_handler = handler;
+}
diff --git a/clients/window.h b/clients/window.h
index f9da8051..1b4285ff 100644
--- a/clients/window.h
+++ b/clients/window.h
@@ -111,6 +111,9 @@ typedef int (*window_motion_handler_t)(struct window *window,
 				       int32_t x, int32_t y,
 				       int32_t sx, int32_t sy, void *data);
 
+typedef void (*display_drag_offer_handler_t)(struct wl_drag_offer *offer,
+					     struct display *display);
+
 struct window *
 window_create(struct display *display, const char *title,
 	      int32_t x, int32_t y, int32_t width, int32_t height);
@@ -149,6 +152,9 @@ window_set_fullscreen(struct window *window, int fullscreen);
 void
 window_set_user_data(struct window *window, void *data);
 
+void *
+window_get_user_data(struct window *window);
+
 void
 window_set_redraw_handler(struct window *window,
 			  window_redraw_handler_t handler);
@@ -191,7 +197,13 @@ window_set_frame_handler(struct window *window,
 			 window_frame_handler_t handler);
 
 void
-window_start_drag(struct window *window, struct input *input, uint32_t time);
+display_set_drag_offer_handler(struct display *display,
+			       display_drag_offer_handler_t handler);
+
+struct wl_drag *
+window_start_drag(struct window *window, struct input *input, uint32_t time,
+		  const struct wl_drag_listener *listener, void *data);
+
 
 void
 input_get_position(struct input *input, int32_t *x, int32_t *y);
diff --git a/compositor.c b/compositor.c
index 9c8d5b5b..32f9d11b 100644
--- a/compositor.c
+++ b/compositor.c
@@ -573,9 +573,49 @@ shell_resize(struct wl_client *client, struct wl_shell *shell,
 	wlsc_input_device_set_pointer_image(wd, pointer);
 }
 
+static void
+destroy_drag(struct wl_resource *resource, struct wl_client *client)
+{
+	struct wl_drag *drag =
+		container_of(resource, struct wl_drag, resource);
+
+	/* FIXME: More stuff */
+
+	free(drag);
+}
+
+const static struct wl_drag_interface drag_interface;
+
+static void
+shell_create_drag(struct wl_client *client,
+		  struct wl_shell *shell, uint32_t id)
+{
+	struct wl_display *display = wl_client_get_display(client);
+	struct wl_drag *drag;
+
+	drag = malloc(sizeof *drag);
+	if (drag == NULL) {
+		wl_client_post_event(client,
+				     (struct wl_object *) display,
+				     WL_DISPLAY_NO_MEMORY);
+		return;
+	}
+
+	memset(drag, 0, sizeof *drag);
+	drag->resource.base.id = id;
+	drag->resource.base.interface = &wl_drag_interface;
+	drag->resource.base.implementation =
+		(void (**)(void)) &drag_interface;
+
+	drag->resource.destroy = destroy_drag;
+
+	wl_client_add_resource(client, &drag->resource);
+}
+
 const static struct wl_shell_interface shell_interface = {
 	shell_move,
-	shell_resize
+	shell_resize,
+	shell_create_drag
 };
 
 static void
@@ -589,7 +629,7 @@ compositor_create_surface(struct wl_client *client,
 	if (surface == NULL) {
 		wl_client_post_event(client,
 				     (struct wl_object *) ec->wl_display,
-				     WL_DISPLAY_NO_MEMORY, 0);
+				     WL_DISPLAY_NO_MEMORY);
 		return;
 	}
 
@@ -695,8 +735,6 @@ pick_surface(struct wlsc_input_device *device, int32_t *sx, int32_t *sy)
 	return NULL;
 }
 
-static void
-wl_drag_reset(struct wl_drag *drag);
 static void
 wl_drag_set_pointer_focus(struct wl_drag *drag,
 			  struct wlsc_surface *surface, uint32_t time,
@@ -797,11 +835,12 @@ notify_motion(struct wlsc_input_device *device, uint32_t time, int x, int y)
 
 	case WLSC_DEVICE_GRAB_DRAG:
 		es = pick_surface(device, &sx, &sy);
-		wl_drag_set_pointer_focus(&device->drag,
+		wl_drag_set_pointer_focus(device->drag,
 					  es, time, x, y, sx, sy);
 		if (es)
-			wl_surface_post_event(&es->base, &device->drag.base,
-					      WL_DRAG_MOTION,
+			wl_surface_post_event(&es->base,
+					      &device->drag->drag_offer.base,
+					      WL_DRAG_OFFER_MOTION,
 					      time, x, y, sx, sy);
 		break;
 
@@ -817,7 +856,7 @@ notify_motion(struct wlsc_input_device *device, uint32_t time, int x, int y)
 static void
 wlsc_input_device_end_grab(struct wlsc_input_device *device, uint32_t time)
 {
-	struct wl_drag *drag = &device->drag;
+	struct wl_drag *drag = device->drag;
 	struct wlsc_surface *es;
 	int32_t sx, sy;
 
@@ -825,9 +864,10 @@ wlsc_input_device_end_grab(struct wlsc_input_device *device, uint32_t time)
 	case WLSC_DEVICE_GRAB_DRAG:
 		if (drag->target)
 			wl_client_post_event(drag->target,
-					     &drag->base, WL_DRAG_DROP);
+					     &drag->drag_offer.base,
+					     WL_DRAG_OFFER_DROP);
 		wl_drag_set_pointer_focus(drag, NULL, time, 0, 0, 0, 0);
-		wl_drag_reset(drag);
+		device->drag = NULL;
 		break;
 	default:
 		break;
@@ -1012,80 +1052,89 @@ wl_drag_set_pointer_focus(struct wl_drag *drag,
 	if (drag->pointer_focus &&
 	    (!surface || drag->pointer_focus->client != surface->base.client))
 		wl_surface_post_event(drag->pointer_focus,
-				      &drag->base,
-				      WL_DRAG_POINTER_FOCUS,
+				      &drag->drag_offer.base,
+				      WL_DRAG_OFFER_POINTER_FOCUS,
 				      time, NULL, 0, 0, 0, 0);
-	if (surface) {
+
+	if (surface &&
+	    (!drag->pointer_focus ||
+	     drag->pointer_focus->client != surface->base.client)) {
 		wl_surface_post_event(&surface->base,
-				      &drag->base,
-				      WL_DRAG_POINTER_FOCUS,
-				      time, &surface->base,
-				      x, y, sx, sy);
+				      (struct wl_object *) surface->compositor->wl_display,
+				      WL_DISPLAY_GLOBAL,
+				      &drag->drag_offer.base,
+				      drag->drag_offer.base.interface->name,
+				      drag->drag_offer.base.interface->version);
 
 		end = drag->types.data + drag->types.size;
 		for (p = drag->types.data; p < end; p++)
 			wl_surface_post_event(&surface->base,
-					      &drag->base,
-					      WL_DRAG_OFFER, *p);
+					      &drag->drag_offer.base,
+					      WL_DRAG_OFFER_OFFER, *p);
 	}
 
+	if (surface) {
+		wl_surface_post_event(&surface->base,
+				      &drag->drag_offer.base,
+				      WL_DRAG_OFFER_POINTER_FOCUS,
+				      time, &surface->base,
+				      x, y, sx, sy);
+
+	}
 
 	drag->pointer_focus = &surface->base;
+	drag->pointer_focus_time = time;
 	drag->target = NULL;
 }
 
 static void
-wl_drag_reset(struct wl_drag *drag)
+drag_offer_accept(struct wl_client *client,
+		  struct wl_drag_offer *offer, uint32_t time, const char *type)
 {
+	struct wl_drag *drag = container_of(offer, struct wl_drag, drag_offer);
 	char **p, **end;
 
+	/* If the client responds to drag pointer_focus or motion
+	 * events after the pointer has left the surface, we just
+	 * discard the accept requests.  The drag source just won't
+	 * get the corresponding 'target' events and eventually the
+	 * next surface/root will start sending events. */
+	if (time < drag->pointer_focus_time)
+		return;
+
+	drag->target = client;
+	drag->type = NULL;
 	end = drag->types.data + drag->types.size;
 	for (p = drag->types.data; p < end; p++)
-		free(*p);
-	wl_array_release(&drag->types);
-	wl_array_init(&drag->types);
-
-	/* FIXME: We need to reset drag->target too, but can't right
-	 * now because we need it for send/drop.
-	 *
-	 * drag->target = NULL;
-	 * drag->source = NULL;
-	 */
-	drag->time = 0;
-	drag->pointer_focus = NULL;
+		if (type && strcmp(*p, type) == 0)
+			drag->type = *p;
+
+	wl_surface_post_event(drag->source, &drag->resource.base,
+			      WL_DRAG_TARGET, drag->type);
 }
 
 static void
-drag_prepare(struct wl_client *client,
-	     struct wl_drag *drag, struct wl_surface *surface, uint32_t time)
+drag_offer_receive(struct wl_client *client,
+		   struct wl_drag_offer *offer, int fd)
 {
-	struct wlsc_input_device *device =
-		(struct wlsc_input_device *) drag->input_device;
-
-	if (&device->pointer_focus->base != surface ||
-	    device->grab_time != time)
-		return;
+	struct wl_drag *drag = container_of(offer, struct wl_drag, drag_offer);
 
-	wl_drag_reset(drag);
-	drag->source = surface;
-	drag->time = time;
+	wl_surface_post_event(drag->source, &drag->resource.base,
+			      WL_DRAG_FINISH, fd);
+	close(fd);
 }
 
+static const struct wl_drag_offer_interface drag_offer_interface = {
+	drag_offer_accept,
+	drag_offer_receive
+};
+
 static void
 drag_offer(struct wl_client *client, struct wl_drag *drag, const char *type)
 {
 	struct wl_display *display = wl_client_get_display (client);
-	struct wlsc_input_device *device =
-		(struct wlsc_input_device *) drag->input_device;
 	char **p;
 
-	if (drag->source == NULL ||
-	    drag->source->client != client ||
-	    device->grab != WLSC_DEVICE_GRAB_MOTION ||
-	    &device->pointer_focus->base != drag->source ||
-	    device->grab_time != drag->time)
-		return;
-
 	p = wl_array_add(&drag->types, sizeof *p);
 	if (p)
 		*p = strdup(type);
@@ -1097,25 +1146,36 @@ drag_offer(struct wl_client *client, struct wl_drag *drag, const char *type)
 
 static void
 drag_activate(struct wl_client *client,
-	      struct wl_drag *drag)
+	      struct wl_drag *drag,
+	      struct wl_surface *surface,
+	      struct wl_input_device *input_device, uint32_t time)
 {
+	struct wl_display *display = wl_client_get_display (client);
 	struct wlsc_input_device *device =
-		(struct wlsc_input_device *) drag->input_device;
-	struct wlsc_surface *surface;
+		(struct wlsc_input_device *) input_device;
+	struct wlsc_surface *target;
 	int32_t sx, sy;
 
-	if (drag->source == NULL ||
-	    drag->source->client != client ||
-	    device->grab != WLSC_DEVICE_GRAB_MOTION ||
-	    &device->pointer_focus->base != drag->source ||
-	    device->grab_time != drag->time)
+	if (device->grab != WLSC_DEVICE_GRAB_MOTION ||
+	    &device->pointer_focus->base != surface ||
+	    device->grab_time != time)
 		return;
 
-	wlsc_input_device_start_grab(device, drag->time,
+	drag->source = surface;
+	drag->input_device = input_device;
+
+	drag->drag_offer.base.interface = &wl_drag_offer_interface;
+	drag->drag_offer.base.implementation =
+		(void (**)(void)) &drag_offer_interface;
+
+	wl_display_add_object(display, &drag->drag_offer.base);
+
+	wlsc_input_device_start_grab(device, time,
 				     WLSC_DEVICE_GRAB_DRAG);
 
-	surface = pick_surface(device, &sx, &sy);
-	wl_drag_set_pointer_focus(&device->drag, surface, drag->time,
+	device->drag = drag;
+	target = pick_surface(device, &sx, &sy);
+	wl_drag_set_pointer_focus(device->drag, target, time,
 				  device->x, device->y, sx, sy);
 }
 
@@ -1125,102 +1185,19 @@ drag_cancel(struct wl_client *client, struct wl_drag *drag)
 	struct wlsc_input_device *device =
 		(struct wlsc_input_device *) drag->input_device;
 
-	if (drag->source == NULL ||
-	    drag->source->client != client ||
-	    device->grab != WLSC_DEVICE_GRAB_DRAG)
+	if (drag->source == NULL || drag->source->client != client)
 		return;
 
 	wlsc_input_device_end_grab(device, get_time());
-}
-
-static void
-drag_accept(struct wl_client *client,
-	    struct wl_drag *drag, const char *type)
-{
-	char **p, **end;
-
-	/* If the client responds to drag motion events after the
-	 * pointer has left the surface, we just discard the accept
-	 * requests.  The drag source just won't get the corresponding
-	 * 'target' events and the next surface/root will start
-	 * sending events. */
-	if (drag->pointer_focus == NULL)
-		return;
-
-	/* The accept request may arrive after the pointer has left
-	 * the surface that received the drag motion events that
-	 * triggered the request.  But if the pointer re-enters the
-	 * surface (or another surface from the same client) and we
-	 * receive the 'accept' requests from the first visit to the
-	 * surface later, this check will pass.  Then we end up
-	 * sending the 'target' event for the old 'accept' requests
-	 * events *after* potentially sending 'target' from the root
-	 * surface, out of order.  So we need to track the time of
-	 * pointer_focus and this request needs a time stamp so we can
-	 * compare and reject 'accept' requests that are older than
-	 * the pointer_focus in timestamp. */
-	if (drag->pointer_focus->client != client)
-		return;
-
-	/* FIXME: We need a serial number here to correlate the accept
-	 * request with a pointer_focus/motion event. */
-	drag->target = client;
-
-	if (type == NULL) {
-		drag->type = NULL;
-	} else {
-		end = drag->types.data + drag->types.size;
-		for (p = drag->types.data; p < end; p++)
-			if (strcmp(*p, type) == 0)
-				drag->type = *p;
-	}
-
-	wl_surface_post_event(drag->source, &drag->base,
-			      WL_DRAG_TARGET, drag->type);
-}
-
-static void
-drag_receive(struct wl_client *client,
-	     struct wl_drag *drag, int fd)
-{
-	wl_surface_post_event(drag->source, &drag->base, WL_DRAG_FINISH, fd);
-	close(fd);
+	device->drag = NULL;
 }
 
 static const struct wl_drag_interface drag_interface = {
-	drag_prepare,
 	drag_offer,
 	drag_activate,
 	drag_cancel,
-	drag_accept,
-	drag_receive
 };
 
-static void
-wl_drag_post_device(struct wl_client *client, struct wl_object *global)
-{
-	struct wl_drag *drag = container_of(global, struct wl_drag, base);
-
-	wl_client_post_event(client, global,
-			     WL_DRAG_DEVICE, drag->input_device);
-}
-
-static void
-wl_drag_init(struct wl_drag *drag,
-	     struct wl_display *display, struct wl_input_device *input_device)
-{
-	drag->base.interface = &wl_drag_interface;
-	drag->base.implementation = (void (**)(void))
-		&drag_interface;
-
-	wl_display_add_object(display, &drag->base);
-	wl_display_add_global(display, &drag->base, wl_drag_post_device);
-
-	drag->source = NULL;
-	wl_array_init(&drag->types);
-	drag->input_device = input_device;
-}
-
 void
 wlsc_input_device_init(struct wlsc_input_device *device,
 		       struct wlsc_compositor *ec)
@@ -1231,8 +1208,6 @@ wlsc_input_device_init(struct wlsc_input_device *device,
 	wl_display_add_object(ec->wl_display, &device->base.base);
 	wl_display_add_global(ec->wl_display, &device->base.base, NULL);
 
-	wl_drag_init(&device->drag, ec->wl_display, &device->base);
-
 	device->x = 100;
 	device->y = 100;
 	device->ec = ec;
diff --git a/compositor.h b/compositor.h
index 44fc1a8d..097cdc64 100644
--- a/compositor.h
+++ b/compositor.h
@@ -107,7 +107,7 @@ struct wlsc_input_device {
 	int32_t grab_width, grab_height;
 	int32_t grab_dx, grab_dy;
 	uint32_t grab_button;
-	struct wl_drag drag;
+	struct wl_drag *drag;
 
 	struct wlsc_listener listener;
 };
diff --git a/protocol.xml b/protocol.xml
index 390fb232..0f2bd98d 100644
--- a/protocol.xml
+++ b/protocol.xml
@@ -66,7 +66,7 @@
   </interface>
 
   <interface name="buffer" version="1">
-    <request name="destroy"/>
+    <request name="destroy" type="destructor"/>
   </interface>
 
   <interface name="shell" version="1">
@@ -84,6 +84,10 @@
       <arg name="edges" type="uint"/>
     </request>
 
+    <request name="create_drag">
+      <arg name="id" type="new_id" interface="drag"/>
+    </request>
+
     <event name="configure">
       <arg name="time" type="uint"/>
       <!-- Same as edges except also move (16) -->
@@ -97,46 +101,65 @@
   </interface>
 
   <interface name="drag" version="1">
-    <request name="prepare">
-      <!-- Start a drag action from given surface and device for the
-	   grab started by the button click at time -->
-      <arg name="surface" type="object" interface="surface"/>
-      <arg name="time" type="uint"/>
-    </request>
-
     <!-- Add an offered mime type.  Can be called several times to
          offer multiple types, but must be called before 'activate'. -->
     <request name="offer">
       <arg name="type" type="string"/>
     </request>
 
-    <request name="activate"/>
+    <request name="activate">
+      <arg name="surface" type="object" interface="surface"/>
+      <arg name="input_device" type="object" interface="input_device"/>
+      <arg name="time" type="uint"/>
+    </request>
 
-    <!-- Cancel the drag. -->
-    <request name="cancel"/>
+    <!-- Destroy the drag and cancel the session. -->
+    <request name="destroy" type="destructor"/>
+
+    <!-- Sent when a target accepts pointer_focus or motion events.
+         If a target does not accept any of the offered types, type is
+         NULL -->
+    <event name="target">
+      <arg name="mime_type" type="string"/>
+    </event>
+
+    <!-- Sent when the drag is finished.  The final mime type is that
+         of the last target event.  If that was NULL, no drag target
+         was valid when the drag finished, fd is undefined and the
+         source should not send data.  The event is also sent in case
+         a drag source tries to activate a drag after the grab was
+         released, in which case mime_type will also be NULL. -->
+    <event name="finish">
+      <arg name="fd" type="fd"/>
+    </event>
+  </interface>
 
-    <!-- Called by the drag target to accept the offer of the given
-         type -->
+
+  <interface name="drag_offer" version="1">
+    <!-- Call to accept the offer of the given type -->
     <request name="accept">
+      <arg name="time" type="uint"/>
       <arg name="type" type="string"/>
     </request>
 
-    <!-- Called by the drag target to initiate the drag finish
-         sequence.  Send the pipe fd to the compositor, which forwards
-         it to the source in the 'finish' event -->
+    <!-- Called to initiate the drag finish sequence.  Sends the pipe
+         fd to the compositor, which forwards it to the source in the
+         'finish' event -->
     <request name="receive">
       <arg name="fd" type="fd"/>
     </request>
 
-    <!-- Sent at connect time to announce the association -->
-    <event name="device">
-      <arg name="device" type="object" interface="input_device"/>
+    <!-- Sent before the pointer_focus event to announce the types
+         offered.  One event per offered mime type. -->
+    <event name="offer">
+      <arg name="type" type="string"/>
     </event>
 
-    <!-- Similar to device::pointer_focus.  Sent to potential
-	 target surfaces to offer drag data.  If the device
-	 leaves the window, the drag stops or the originator cancels
-	 the drag, this event is sent with the NULL surface. -->
+    <!-- Similar to device::pointer_focus.  Sent to potential target
+         surfaces to offer drag data.  If the device leaves the
+         window, the drag stops or the originator cancels the drag,
+         this event is sent with the NULL surface, at which point the
+         drag object may no longer be valid. -->
     <event name="pointer_focus">
       <arg name="time" type="uint"/>
       <arg name="surface" type="object" interface="surface"/>
@@ -146,12 +169,6 @@
       <arg name="surface_y" type="int"/>
     </event>
 
-    <!-- Sent after the pointer_focus event to announce the types
-	 offered.  One event per offered mime type. -->
-    <event name="offer">
-      <arg name="type" type="string"/>
-    </event>
-
     <!-- Similar to device::motion.  Sent to potential target surfaces
 	 as the drag pointer moves around in the surface. -->
     <event name="motion">
@@ -162,34 +179,15 @@
       <arg name="surface_y" type="int"/>
     </event>
 
-    <!-- Sent to drag originator in response to pointer_focus and
-         motion events.  If a target does not accept any of the
-         offered types, type is NULL -->
-    <event name="target">
-      <arg name="mime_type" type="string"/>
-    </event>
-
-    <!-- Sent to target, to indicate that the drag is finishing.  The
-         last motion/pointer_focus event gives the location of the
-         drop.  Target must respond with the 'receive' request, which
-         sends an fd to the source for writing the drag data. -->
+    <!-- Sent to indicate that the drag is finishing.  The last
+         motion/pointer_focus event gives the location of the drop.
+         Target must respond with the 'receive' request, which sends
+         an fd to the source for writing the drag data. -->
     <event name="drop"/>
-
-    <!-- Sent to drag source when the drag is finished.  The final
-         mime type is that of the last target event.  If that was
-         NULL, no drag target was valid when the drag finished, fd is
-         undefined and the source should not send data.  The event is
-         also sent in case a drag source tries to activate a drag
-         after the grab was released, in which case mime_type will
-         also be NULL. -->
-    <event name="finish">
-      <arg name="fd" type="fd"/>
-    </event>
-
   </interface>
 
   <interface name="surface" version="1">
-    <request name="destroy"/>
+    <request name="destroy" type="destructor"/>
 
     <request name="attach">
       <arg name="buffer" type="object" interface="buffer"/>
diff --git a/scanner.c b/scanner.c
index b4d76681..613e12b9 100644
--- a/scanner.c
+++ b/scanner.c
@@ -74,6 +74,7 @@ struct message {
 	char *uppercase_name;
 	struct wl_list arg_list;
 	struct wl_list link;
+	int destructor;
 };
 
 enum arg_type {
@@ -177,6 +178,11 @@ start_element(void *data, const char *element_name, const char **atts)
 			wl_list_insert(ctx->interface->event_list.prev,
 				       &message->link);
 
+		if (type != NULL && strcmp(type, "destructor") == 0)
+			message->destructor = 1;
+		else
+			message->destructor = 0;
+
 		ctx->message = message;
 	} else if (strcmp(element_name, "arg") == 0) {
 		arg = malloc(sizeof *arg);
@@ -263,6 +269,7 @@ emit_stubs(struct wl_list *message_list, struct interface *interface)
 {
 	struct message *m;
 	struct arg *a, *ret;
+	int has_destructor, has_destroy;
 
 	/* We provide a hand written constructor for the display object */
 	if (strcmp(interface->name, "display") != 0)
@@ -293,6 +300,33 @@ emit_stubs(struct wl_list *message_list, struct interface *interface)
 	       interface->name, interface->name, interface->name,
 	       interface->name);
 
+	has_destructor = 0;
+	has_destroy = 0;
+	wl_list_for_each(m, message_list, link) {
+		if (m->destructor)
+			has_destructor = 1;
+		if (strcmp(m->name, "destroy)") == 0)
+			has_destroy = 1;
+	}
+
+	if (!has_destructor && has_destroy) {
+		fprintf(stderr,
+			"interface %s has method named destroy but"
+			"no destructor", interface->name);
+		exit(EXIT_FAILURE);
+	}
+
+	/* And we have a hand-written display destructor */
+	if (!has_destructor && strcmp(interface->name, "display") != 0)
+		printf("static inline void\n"
+		       "wl_%s_destroy(struct wl_%s *%s)\n"
+		       "{\n"
+		       "\twl_proxy_destroy("
+		       "(struct wl_proxy *) %s);\n"
+		       "}\n\n",
+		       interface->name, interface->name, interface->name,
+		       interface->name);
+
 	if (wl_list_empty(message_list))
 		return;
 
@@ -347,6 +381,11 @@ emit_stubs(struct wl_list *message_list, struct interface *interface)
 		}
 		printf(");\n");
 
+		if (m->destructor)
+			printf("\n\twl_proxy_destroy("
+			       "(struct wl_proxy *) %s);\n",
+			       interface->name);
+
 		if (ret)
 			printf("\n\treturn (struct wl_%s *) %s;\n",
 			       ret->interface_name, ret->name);
@@ -442,6 +481,9 @@ static const char client_prototypes[] =
 	"wl_proxy_create_for_id(struct wl_display *display,\n"
 	"\t\t       const struct wl_interface *interface, uint32_t id);\n"
 
+	"extern void\n"
+	"wl_proxy_destroy(struct wl_proxy *proxy);\n\n"
+
 	"extern int\n"
 	"wl_proxy_add_listener(struct wl_proxy *proxy,\n"
 	"\t\t      void (**implementation)(void), void *data);\n\n"
diff --git a/wayland-client.c b/wayland-client.c
index 47df184d..dd0500f5 100644
--- a/wayland-client.c
+++ b/wayland-client.c
@@ -146,6 +146,18 @@ wl_proxy_create(struct wl_proxy *factory,
 				      wl_display_allocate_id(factory->display));
 }
 
+WL_EXPORT void
+wl_proxy_destroy(struct wl_proxy *proxy)
+{
+	struct wl_listener *listener, *next;
+
+	wl_list_for_each_safe(listener, next, &proxy->listener_list, link)
+		free(listener);
+
+	wl_hash_table_remove(proxy->display->objects, proxy->base.id);
+	free(proxy);
+}
+
 WL_EXPORT int
 wl_proxy_add_listener(struct wl_proxy *proxy,
 		      void (**implementation)(void), void *data)
diff --git a/wayland-server.h b/wayland-server.h
index acfe7914..4292be05 100644
--- a/wayland-server.h
+++ b/wayland-server.h
@@ -124,8 +124,13 @@ struct wl_visual {
 	struct wl_object base;
 };
 
-struct wl_drag {
+struct wl_drag_offer {
 	struct wl_object base;
+};
+
+struct wl_drag {
+	struct wl_resource resource;
+	struct wl_drag_offer drag_offer;
 	struct wl_surface *source;
 	struct wl_surface *pointer_focus;
 	struct wl_client *target;
@@ -133,7 +138,7 @@ struct wl_drag {
 	struct wl_input_device *input_device;
 	struct wl_array types;
 	const char *type;
-	uint32_t time;
+	uint32_t pointer_focus_time;
 };
 
 void