Open a 50 GB archive with the memory of its central directory
openZipRange() (1.1) never fetches: you inject a ByteRangeSource —
{ size, read(offset, length) } — and the engine asks for exactly the bytes a read
needs: one tail window, the central directory, then the local header and payload of the
entry you ask for. Every limit and cross-check of openZip() applies. Below,
the source counts its reads so you can see the arithmetic.
1 · A dropped file, read by ranges
Drop a large .zip (a JAR, a dataset, a game asset bundle). The source is
Blob.slice() — the file is not loaded into memory — and the counter shows how
little of it the inventory needs.
or click to choose a file — read by ranges, never uploaded, never fully loaded
2 · A URL served with HTTP Range
Any URL that answers HEAD with a Content-Length, honours
Range (status 206) and allows this origin through CORS — object storage with a
public bucket, a CDN, GitHub Pages. The source is fetch with a
Range: bytes=… header.
Click a row to read that entry — by ranges, CRC-verified — and watch the counter.
| Name | Method | Uncompressed | Compressed | Offset |
|---|