13. Feasibility Assessment

13.1 Overall Assessment

The design is feasible because vendor DSP, gfortran, fftw3f, FastAPI, rigctld and PortAudio are available. The largest risks are Improved decoder deadline, Fortran/OpenMP isolation, audio/clock correctness and public TX safety. The vertical-slice order attacks these before UI polish.

13.2 Risks

ID Risk Mitigation / Evidence
R1 macOS/Linux Fortran/OpenMP build/link incompatibility M1 minimal source build, exact four-export ABI smoke test, generated ELF map regression, documented toolchain
R2 Improved profile misses TX decision deadline Profile 3 default, Auto threads, latency histogram, late result display-only
R3 packjt77/ft8var mutable state races One worker owner, binding lock, exact OpenMP team check, thread-private slot/FFT/cycle state, deterministic A8 owner, batched Fortran return and repeated parallel-region regression
R4 Worker crash or native fault Separate process, generation-tagged IPC, fail-closed stop/restart
R5 Audio device loss, overrun or wrong rate One audio owner, explicit assertions, fault injection, real-radio acceptance
R6 UTC step/drift causes invalid slot transmission Epoch slot math, clock health interlock, manual re-arm
R7 Public attacker or concurrent client starts TX Caddy TLS, strong auth, Origin/Host, rate limits, one lease, universal STOP
R8 Browser backpressure delays safety traffic Separate bounded streams; waterfall drops only
R9 macOS background service cannot access audio User LaunchAgent and explicit permission acceptance
R10 Diagnostic bundle leaks sensitive local data Password re-auth, prominent raw-data warning, user-authorized local export

Task 5’s direct-A8 hook repeatedly decodes a clean fixed message and the atomic near-Rx gate has a native threshold/concurrency probe. A deterministic weak direct-A8 fixture could not be reproduced across fresh library processes, so that weak direct-A8 fixture remains a strict expected failure and a documented test-evidence limitation. It is not an architecture blocker; it must not be misrepresented as weak-signal sensitivity coverage.

13.3 Assumptions

ID Assumption
A1 Operator owns a domain whose DNS points to the Caddy host
A2 FT-710 and supported host expose stable 48 kHz USB audio
A3 rigctld supports required frequency/mode/PTT operations for the target radio
A4 Host clock is normally synchronized closely enough for FT8; unsafe state is detectable
A5 Apple Silicon/macOS has sufficient CPU for profile 3 Improved decoding with OpenMP
A6 One personal station and one simultaneous QSO are sufficient for V1.0

13.4 Resolved Issues

13.5 Open Issues

ID Issue When resolved
I10 FT-710 PTT measured (2026-08-01, Mac mini M2, C-Media USB audio, 5 W): priority STOP releases PTT in 0.26–0.31 s; a 2 s CAT-keyed transmission closes 0.8–0.9 s after audio ends. Residual uncertainty: audio-lead at key-down not yet instrumented macOS acceptance
I11 Linux distributions and audio backend matrix beyond baseline After V1.0

Open issues may tune implementation parameters but cannot weaken the fixed ownership, rate, lease or safety decisions.

13.6 Go/No-go Gates