Adjacent approaches and escape hatches
Last reviewed: September 30, 2026.
These options widen the design space without being interchangeable with the four main backends. They have not been implemented or benchmarked in this research.
| Approach | What it changes | Fit for custom terminal | Principal cost |
|---|---|---|---|
| QEMU system emulation / TCG | Emulates guest CPU and devices without requiring usable EL2 virtualization | Own emulator lifecycle and guest PTY transport; full Linux possible | CPU, battery and boot/memory overhead; substantial native packaging |
| SSH to remote NixOS | Moves store, builds and kernel services off device | Existing terminal can render a remote PTY | Connectivity, latency, host administration and offline limitations |
| Remote Nix builders | Offloads build execution | Useful adjunct to any local Nix client | Local result still needs a compatible runtime and logical store |
| Bionic-native Nix / package universe | Targets Android libc and filesystem/security model | Native app integration potentially simpler | Separate platform/package coverage; ordinary glibc Linux outputs are not automatically compatible |
| Root bind/chroot helper | Supplies canonical store without PRoot | Owned privilege broker plus native PTY | Root policy, UID separation and Android kernel limitations |
QEMU's system emulation documentation distinguishes full system emulation and accelerators; do not confuse PRoot with CPU emulation. A native crosvm/QEMU wrapper seen in DroidVM is architectural prior art, not permission to assume a particular accelerator is available.
Nix's distributed build model solves build placement, not local loader compatibility. The ELF example demonstrates why a successful remote build does not make its interpreter path appear on Android.
A Bionic port would need explicit packaging and cache policy, loader assumptions, dependency coverage and store mapping. This survey found no locally demonstrated drop-in Bionic replacement for the inspected nix-on-droid Linux bootstrap. Treat it as a separate portability project rather than a near-term substitute.