max / quasi

Destination::Back, so a key, a Close and a gesture mean one thing Decided in quasicoherent 33c27e81 (Max chose a). A route cannot name back: which route it is depends on history the runtime holds and the description does not. So it is a destination, where "not a route" already lives, and Action::back() is the one way to say it -- a keyboard binding, a visible Close and a touch host's system gesture all carry the same thing and cannot disagree about where back goes. Whose history is the part worth stating and the doc states it: the runtime's. Each host keeps what it has been asked for and answers this by popping it. A webview maps that onto the browser's history, not the other way round. is_local stays false for it. Local says no request is made at all, and going back usually makes one -- what this variant says is that the address is the host's to supply, which is a different fact. Both hosts handle it before their outside-the-app guard, which reads route() as "None, therefore elsewhere" and would hand the host an empty address to open. The webview performs it rather than marking it, because there the browser is the host holding the history. `call history.back()` and not a js block, so it runs under script-src 'self'. The gesture is a parameter: a Chrome binding's button is hidden and never clicked, so a program listening only for a click would leave every bound key silently doing nothing. Its filter is interpolated, and it is this crate's own validated text -- modifiers are literals and the base is one escaped character or a name from a closed list -- which is the discipline the handle gate already applies to a region id. A back binding gets no hx-target either: nothing is coming back to put anywhere. A Goto carrying Back stays a description bug, like Goto(Local), and the redirect doc now says why: no header says "back", the retained-screen hosts drop it too, and the three agreeing matters more than one of them being clever. What a description says instead is an action on a control.