Implementation Language Selection

— — Note

Scope: Evaluates Go vs. Python specifically against the agent's actual operational requirements — a small, infrequently-run, I/O-bound client that extracts structured data and streams it to a remote endpoint.

Context

The existing implementation is Python 3, packaged as a single self-extracting frozen executable.

The redesigned agent (Note 0002) introduces a requirement the existing implementation does not have: asymmetric cryptographic signing (Ed25519) of requests to the platform server.

Evaluation Axes

STEP— Dependency Management (Primary Factor)

Go: Standard library includes production-grade crypto/ed25519 with zero external dependencies. Python: Standard library covers only symmetric primitives. Asymmetric signing requires compiled native C extensions (cryptography package), re-introducing native bundling complexity.

STEP— Distribution & Antivirus Scrutiny

Go: Compiles to a single static native binary. Python: Frozen executables self-extract to temporary directories on every launch, triggering endpoint heuristic & antivirus quarantine flags on Windows.

STEP— Cross-Compilation

Go: Native cross-compilation from any OS to Windows target (GOOS=windows GOARCH=amd64). Python: Requires a dedicated Windows build host or complex cross-build toolchain.

STEP— Error Handling & Unattended Execution

Go: Explicit error propagation prevents accidental silent exception suppression. Python: Broad except Exception: blocks in legacy codebase led to silent failures.

Non-Factors

Concurrency: Neutral. Workload is I/O-bound and low volume per run.
XML/Data Parsing: Neutral. Malformed control characters in raw exports break standard parsers in both languages, requiring custom tolerant tokenizers.

Decision

DEC — Decision · ACCEPTEDaccepted

Implement the redesigned desktop agent in Go.

package agent

import (
	"crypto/ed25519"
	"encoding/hex"
)

// SignRequest signs an outgoing HTTP request payload using the local private key
func SignRequest(privateKey ed25519.PrivateKey, payload []byte) string {
	signature := ed25519.Sign(privateKey, payload)
	return hex.EncodeToString(signature)
}

Consequences

TRD — Trade-off

Gain

Zero-dependency Ed25519 signing from stdlib; single static binary reduces antivirus false positives and deployment friction.

Cost

Requires a full rewrite of the desktop agent (Python and Go are not source-compatible).

Single static binary distribution simplifies deployment and code-signing.
Zero external native dependencies for Ed25519 request signing.
Cross-compilation simplicity for target Windows environments.

Alternatives Considered

Python with PyInstaller / cx_Freeze: Rejected due to native C extension bundling complexity for Ed25519 and antivirus false positives.
Rust: Rejected as unnecessary overhead for an I/O-bound orchestration client with no zero-cost abstraction needs.
Go: Accepted for standard library cryptography, static binaries, and clean cross-compilation.