3 - generate pointer_focus on raise/lower, move windows, all kinds of
4 changes in surface stacking.
6 - don't store globals on client side, require global_handler like
11 - dnd, figure out large object transfer: through wayland protocol or
12 pass an fd through the compositor to the other client and let them
15 - copy-n-paste, store data in server (only one mime-type available)
16 or do X style (content mime-type negotiation, but data goes away
19 - protocol for setting the cursor image
21 - should we have a mechanism to attach surface to cursor for
22 guaranteed non-laggy drag?
24 - drawing cursors, moving them, cursor themes, attaching surfaces
25 to cursors. How do you change cursors when you mouse over a
26 text field if you don't have subwindows? This is what we do: a
27 client can set a cursor for a surface and wayland will set that
28 on enter and revert to default on leave
30 - Discard buffer, as in "wayland discarded your buffer, it's no
31 longer visible, you can stop updating it now.", reattach, as in "oh
32 hey, I'm about to show your buffer that I threw away, what was it
33 again?". for wayland system compositor vt switcing, for example,
34 to be able to throw away the surfaces in the session we're
35 switching away from. for minimized windows that we don't want live
38 - Consolidate drm buffer upload with a create_buffer request, returns
39 buffer object we can use in surface.attach, cache.upload and
40 input.attach? Will increase object id usage significantly, each
41 buffer swap allocates and throws away a new id. Does consolidate
42 the details of a buffer very nicely though.
44 compositor.create_buffer(new_id, visual, name, stride, width, height)
46 surface.attach(buffer)
48 cache.upload(buffer, x, y, width, height)
50 input.set_cursor(buffer, hotspot_x, hotspot_y)
52 Doesn't increase id usage too much, can keep buffers around.
54 - Move/resize protocol in the style of the dnd protocol: a surface
55 who has a grabbed device can send a request to initiate a
56 resize(top/bottom+rigth/left) or a move. The compositor will then
57 resize or move the window and take into account windows, panels and
58 screen edges and constrain and snap the motion accordingly. As the
59 cursor moves, the compositor sends resize or move (maybe not move
60 events?) events to the app, which responds by attaching a new
61 surface at the new size (optionally, reducing the allocated space
62 to satisfy aspect ratio or resize increments).
64 - Initial placement of surfaces. Guess we can do, 1)
65 surface-relative (menus), 2) pointer-relative (tooltips and
66 right-click menus) or 3) server-decides (all other top-levels).
68 - When a surface is the size of the screen and on top, we can set the
69 scanout buffer to that surface directly. Like compiz unredirect
70 top-level window feature. Except it won't have any protocol state
71 side-effects and the client that owns the surface won't know. We
72 lose control of updates. Should work well for X server root window
73 under wayland. Should be possible for yuv overlays as well.
75 - what about cursors then? maybe use hw cursors if the cursor
76 satisfies hw limitations (64x64, only one cursor), switch to
77 composited cursors if not.
79 - clients needs to allocate the surface to be suitable for
80 scanout, which they can do whenever they go fullscreen.
82 - multihead, screen geometry and crtc layout protocol, hotplug
84 - input device discovery, hotplug
86 - Advertise axes as part of the discovery, use something like
87 "org.wayland.input.x" to identify the axes.
89 - keyboard state, layout events at connect time and when it
90 changes, keyboard leds
96 - synaptics, 3-button emulation, scim
98 - sparse/gcc plugin based idl compiler
102 - xml based description instead?
104 - Figure out if we need the batch/commit scheme and what to do
105 instead. Since dropping the "copy" request, we have a race between
106 copy from back to front and reporting damage. "copy" did this
107 atomically, but copy is a rendering operation (wayland doesn't do
108 rendering) and requires synchronization between server and client
109 before client can reuse backbuffer.
111 The race condition happens when a client copies new content into
112 its window and then, before the client reports the damage, the
113 compositor then does a partial repaint (triggered by another
114 client) that only pulls in part of the repainted area. It's only a
115 one-frame glitch, as the client will submit the damage and the
116 compositor will repaint the damaged area next frame. And ideally
117 clients should do all rendering as early in the frame as possible
120 - auth; We need to generate a random socket name and advertise that
121 on dbus along with a connection cookie. Something like a method
122 that returns the socket name and a connection cookie. The
123 connection cookie is just another random string that the client
124 must pass to the wayland server to become authenticated. The
125 Wayland server generates the cookie on demand when the dbus method
126 is called and expires it after 5s or so.
128 - or just pass the fd over dbus
130 - drm bo access control, authentication, flink_to
132 - Range protocol may not be sufficient... if a server cycles through
133 2^32 object IDs we don't have a way to handle wrapping. And since
134 we hand out a range of 256 IDs to each new clients, we're just
135 talking about 2^24 clients. That's 31 years with a new client
136 every minute... Maybe just use bigger ranges, then it's feasible
137 to track and garbage collect them when a client dies.
139 - Add protocol to let applications specify the effective/logical
140 surface rectangle, that is, the edge of the window, ignoring drop
141 shadows and other padding. The compositor needs this for snapping
142 and constraining window motion. Also, maybe communicate the opaque
143 region of the window (or just a conservative, simple estimate), to
144 let the compositor reduce overdraw.
146 - multi gpu, needs queue and seqno to wait on in requests
152 - eek, so much X legacy stuff there...
154 - draw window decorations in gtkwindow.c
156 - start from alexl's client-side-windows branch
158 - Details about pointer grabs. wayland doesn't have active grabs,
159 menus will behave subtly different. Under X, clicking a menu
160 open grabs the pointer and clicking outside the window pops down
161 the menu and swallows the click. without active grabs we can't
162 swallow the click. I'm sure there much more...
164 - Port Qt? There's already talk about this on the list.
168 - move most of the code from xf86-video-intel into a Xorg wayland
171 - don't ask KMS for available output and modes, use the info from
172 the wayland server. then stop mooching off of drmmode.c.
174 - map multiple wayland input devices to MPX in Xorg.
176 - rootless; avoid allocating and setting the front buffer, draw
177 window decorations in the X server (!), how to map input?
179 - gnome-shell as a wayland session compositor
181 - runs as a client of the wayland session compositor, uses
182 clutter+egl on wayland
184 - talks to an Xorg server as the compositing and window manager
185 for that server and renders the output to a wayland surface.
186 the Xorg server should be modified to take input from the system
187 compositor through gnome-shell, but not allocate a front buffer.
189 - make gnome-shell itself a nested wayland server and allow native
190 wayland clients to connect and can native wayland windows with
191 the windows from the X server.
193 - qemu as a wayland client; session surface as X case
195 - qemu has too simple acceleration, so a Wayland backend like the
196 SDL/VNC ones it has now is trivial.
198 - paravirt: forward wayland screen info as mmio, expose gem ioctls as mmio
200 - mapping vmem is tricky, should try to only use ioctl (pwrite+pread)
202 - not useful for Windows without a windows paravirt driver.
204 - two approaches: 1) do a toplevel qemu window, or 2) expose a
205 wayland server in the guest that forwards to the host wayland
206 server, ie a "remote" compositor, but with the gem buffers
207 shared. could do a wl_connection directly on mmio memory, with
208 head and tail pointers. use an alloc_head register to indicate
209 desired data to write, if it overwrites tail, block guest. just
210 a socket would be easier.
212 - moblin as a wayland compositor
214 - clutter as a wayland compositors