Roadmap
Where deltached is headed.
A small library with a deliberately small surface. This is the direction of travel — grouped by how soon and how sure, not by date. Confidence is highest in Now and lowest in Exploring, and the principles at the end shape every call above.
Nothing in flight
Caught up — the framework wrappers just shipped.
No work is pending right now. What's planned is in Next, below.
Planned
Committed in direction, not yet started.
Scroll-container–aware placement
CoreClamp origin panels to their nearest scroll parent, not just the viewport, so dropdowns sit correctly inside overflowing regions.
RTL-aware origin placement
CoreMirror the directional origin-* growth and viewport clamping for right-to-left layouts.
More persist adapters
PersistFirst-class handling for SVG, video, and gradient / background-image elements, beyond today's text, image, canvas, and surface.
An opt-in dialog companion
EcosystemA separate, optional package for the modal concerns the core deliberately leaves to you — focus trap, scroll lock, Escape, restore-focus — so they aren't rewritten per app.
Research
Ideas under investigation — some of these will never ship.
Gesture-driven dismiss
InteractionHand a pointer's velocity into leave() for drag-to-close that flows out of the same interruptible morph.
View Transitions interop
CoreBridge in-page morphs with the View Transitions API, for shared elements that survive a route change.
List persistence (FLIP)
PersistCarry items across add / remove / reorder within a single layout, not only between two layouts.
A debug overlay
DXA devtool that draws source and target rects and persist layers live — turning persist.debug into something you can see.
Principles & non-goals
What guides every decision above — and what deltached deliberately won't become.
- Core stays measure-and-animate. Focus, scroll lock, and ARIA live in your code or an opt-in package — never bolted onto the core.
- Interruptibility is non-negotiable. Every feature must survive being reversed mid-flight; if it can't, it doesn't ship.
- Framework-agnostic first. Wrappers are thin conveniences over the same vanilla API, never a fork of it.
- Small, with few dependencies. New surface area has to earn its weight.
- 1.0 is about trust, not feature count: a stable API, complete docs, and a tested interruption matrix.