The Stripy Strudel's Journal

When you use ssh to connect to a remote host, it feels like sitting at that machine's keyboard and screen. This illusion is easy to distinguish from reality: in most cases the console window title will contain a clue, and, most important, there are plenty of ways to detach yourself from the remote keyboard and switch to controlling your own, local machine. One of them is to finish the remote session with the logout command. But there are always other ways you can use in case the remote host is “hanging” and you cannot type “logout”. In ssh, for example, it's sufficient to type the ~. sequence (tilde, dot) starting from a new line to force disconnection. This sequence, along with several others, is an escape sequence that is interpreted in a special way. This is important: all characters are usually passed to the remote host as they are, but an escape sequence is handled by the communication medium itself and never reaches the remote server. If one really needs to type this exact sequence on the remote “keyboard”, there is another sequence for this: a double tilde transmits a single tilde character. All of this means that when it comes to tildes, the transparency of ssh breaks, and the illusion of sitting at the remote machine's console becomes incomplete.

An ssh session is, of course, a weak illusion. It's much more natural in case of VNC, Remote Desktop and similar protocols. They let you use the graphical user interface of the remote machine, and in some cases even hear the sounds made by the remote programs. If you switch to the full-screen mode, the illusion becomes almost complete… but the basic principle remains: there are always some escape sequences (at least one to disconnect) that won't reach the remote host but gets handled by the communication medium. It's often some obscure sequence unlikely to be encountered in the normal course of operation, something like Ctrl-Alt-Shift-Esc that you'll hardly hit by accident. That's why those who don't know about the escape sequences have to use the “natural” method of disconnection by closing the session through the remote OS interface. But, of course, there are cases when it's impossible.

This asks for an obvious analogy with virtual reality systems. The simplest case of VR is three-dimensional image on the computer screen, for example, in games. Distinguishing it from reality is trivial: one has to look around. Escape sequences are usually obvious. A VR headgear generates a much better illusion, especially when combined with sensor gloves and other devices reading back the body position. Nevertheless, even such an imitation is easy to recognize, and taking off the headgear is one of the escape sequences. But what if we go further? What if the system imitates three-dimensional image, sound, taste, smell and even the position of the body in space, and does it so realistically that you can't tell it from the reality? What should the escape sequences be like then?

Firstly, the sequences should be obscure enough not to be triggered by accident. Otherwise we would witness people unexpectedly freezing or disappearing when trying to, say, scratch a heel. Secondly, the sequences have to be available in any situation, even when the body motions are seriously limited, or when some body parts are missing — in these situations it's even more likely that someone would want to exit this world. As to where you end up after such a disconnection, this is something we all get to find out even if we never find the secret escape sequences: everyone will eventually exit the natural way, through “logout”.

По-русски: Последовательности для эскапистов