Integration & API Reference

Client Pull & Byte-Range Guide

Recommended flags for high-throughput multi-stream transfers, rapid multiplexed range fetches, and SHA-256 verification.

1. Multi-Stream Bulk & HTTP/2 CLI Pulls

Use persistent HTTP/2 connections for small sparse indexes or aria2c / rclone multi-range segmentation for multi-gigabyte archives.

# Single-socket HTTP/2 pull of the live 64 KiB sample fixture
curl --http2 -fL --retry 3 -o probe-64k.bin \
  https://mirror.mumudada.xyz/artifacts/samples/probe-64k.bin

# 16-connection segmented pull (ideal for multi-GB ONNX / ZIM tarballs)
aria2c --http2=true -x 16 -s 16 -k 1M \
  --checksum=sha-256=76ea02bbe70ceb9805f1987bcdd85a9356b968cef64c58d34f42bb15fff39830 \
  https://mirror.mumudada.xyz/artifacts/samples/probe-64k.bin

# Streaming sync via rclone multi-thread transfers
rclone copyurl https://mirror.mumudada.xyz/v1/catalog.json ./catalog.json --progress

2. Cloud-Native PMTiles HTTP Byte-Range Queries

All objects accept RFC 9110 Range requests over keepalive TLS 1.3 sockets. Geospatial viewers issue rapid batches of 16 KiB to 64 KiB header and directory slice requests against a single persistent connection.

const res = await fetch("/artifacts/samples/probe-64k.bin", {
  headers: { Range: "bytes=0-4095" }
});
const sliceBuffer = await res.arrayBuffer();
console.log("HTTP status:", res.status, "bytes:", sliceBuffer.byteLength);

3. Offline & CI Checksum Verification

Every build synchronized to us-west-lax02 is hashed prior to atomic directory swap. Verify live sample files locally against the signed manifest at /artifacts/checksums/SHA256SUMS.txt.

curl -fsSL https://mirror.mumudada.xyz/artifacts/checksums/SHA256SUMS.txt -o SHA256SUMS.txt
sha256sum -c SHA256SUMS.txt --ignore-missing

5. High-Concurrency & Bulk Egress Fair-Use Policy

Keep-Alive Preferred: Please reuse persistent HTTP/2 TLS sessions when dispatching rapid Range requests or lockfile metadata queries. Opening thousands of un-pooled short-lived TCP handshakes slows client throughput without increasing mirror parallelism.

Automated CI build fleets and academic dataset crawlers may sustain up to 16 parallel byte-range streams per client IP without pre-registration. No cookies, telemetry beacons, or external third-party fonts are served by this node.