Terminal Spec: Synchronized Output

terminal-synchronized-output.md

Synchronized Output

Synchronized output is merely implementing the feature as inspired by iTerm2 synchronized output, except that it's not using the rare DCS but rather the well known SM ? and RM ?. iTerm2 has now also adopted to use the new syntax instead of using DCS.

Semantics

When rendering the screen of the terminal, the Emulator usually iterates through each visible grid cell and renders its current state. With applications updating the screen a at higher frequency this can cause tearing.

This mode attempts to mitigate that.

When the synchronization mode is enabled following render calls will keep rendering the last rendered state. The terminal Emulator keeps processing incoming text and sequences. When the synchronized update mode is disabled again the renderer may fetch the latest screen buffer state again, effectively avoiding the tearing effect by unintentionally rendering in the middle a of an application screen update.

Feature detection

Use CSI ? 2026 $ p to query the state of the (DEC) mode 2026. This works for any private mode number. If you get nothing back (DECRQM not implemented at all) or you get back a CSI ? 2026 ; 0 $ y then synchronized output is not supported. See DECRQM (request) and DECRPM (response) for more details.

DECRPM can respond with different values.

Value	Documentation	Relevance for synchronized output mode ?2026
0	Mode is not recognized	not supported
1	Set	supported and screen updates are not shown to the user until mode is disabled
2	Reset	supported and screen updates are shown as usual (e.g. as soon as they arrive)
3	Permanently set	undefined
4	Permanently reset	not supported

Using the feature

Use CSI ? 2026 h to enable batching output commands into a command queue.

Use CSI ? 2026 l when done with your current frame rendering, implicitly updating the render state by reading out the latest grid buffer state.

Notation

Some developers name the beginning and end of such a synchronized frame (and therefore the instance)

• BSU (begin synchronized update, CSI ? 2026 h), and
• ESU (end synchronized update, CSI ? 2026 l).

Timeout

So far there is no real concensus if and if so how long a timeout should be. The toolkit/application implementer should keep this in mind. However, a too short timeout (maybe due to a very slow connection) won't be worse than having no synchronized output at all.

Adoption State

Support	Terminal/Tookit/App	Notes
n/a	xterm.js	see tracker xterm.js#3375
not yet	Windows Terminal	Proof-of-concept implementation by @j4james exists; tracker: wt#8331
✅	Contour
✅	mintty
✅	Jexer
✅	notcurses	see tracker: notcurses#1582
✅	foot	terminal emulator https://codeberg.org/dnkl/foot
✅	Wezterm	see tracker: wezterm#882
not yet	VTE / gnome-terminal	see tracker: gitlab/vte#15
✅	iTerm2
✅	Kitty	since 5768c54c5b5763e4bbb300726b8ff71b40c128f8
planned	Warp	see tracker: warpdotdev/Warp#2185
✅	Alacritty	since 0.13.0
unknown	Konsole
unknown	urxvt
unknown	st

In case some project is adding support for this feature, please leave a comment or contact me, so we can keep the spec and implementation state table up to date.

It may be useful to clarify the semantics of this in relation to the alternate screen buffer. The semantics I would prefer (x-ref: inducer/pudb#48) would be for this mode to survive buffer switches, so that, upon a "next" command in a debugger, synchronized output can be enabled, a buffer switch can occur, the debuggee be run, and another buffer switch can occur, and the next rendered state be, once again, of the debugger UI.