archive.org /download web serving path (the free CDN)
- Free allowance
- Serving is unmetered and free; no per-file download cap and no egress fee.
- Verdict
- caveatusable, with a named limit or an unprobed read path
- Cost
- free tier
- Identity
- anonymousno card, no phone
- Binding limit
- per-ip-shaping1.19 MB/s; 1.66 MB/s
- Evidence
- reported9 sources; a third party said so
- Freshness
- stale on 2026-11-05checked 2026-10-21 (inherited); stale on 2026-11-05
| Free tier | Serving is unmetered and free; no per-file download cap and no egress fee. Practical ceiling is throughput, not quota. unit mixed; metered on unstated |
|---|---|
| Price | none stated |
| Limits | Canonical URLs 302 at least once (clients need --location-trusted when sending Authorization); do NOT link iaNNNNNN.us.archive.org directly (nodes move); 429/503 under load; no auth on GET, so everything is world-readable. |
| Identity | anonymous no credential is needed to read: an anonymous GET of archive.org/download/<identifier>/<filename> redirects once to a datanode and serves the file. Uploading needs an Internet Archive account whose form asks an email address, a screen name and a password, and asks no phone number |
| Egress | Free / unmetered. Docs: 'high performance downloads are achieved via the archive web infrastructure' — i.e. GET https://archive.org/download/<item>/<file>, which 302s to an iaNNNNNN.us.archive.org datanode. BitTorrent webseeds supplement it. |
| Per-object cap | not stated |
| Retention | not stated |
| Interface | rest, rclone, cli rest |
| Tooling | plain HTTPS GET with Range (measured: 206 + accept-ranges: bytes); ia CLI download w/ --checksum resume; rclone 'internetarchive' reads back through the same front end |
| Composes with | Parallelise across items — different items can be pinned to different iaNNNNNN datanodes, so aggregate throughput scales with item count rather than with connections to one item. |
| Reliability | Throughput is the binding constraint, not quota: measured 1.19 MB/s on one 90s stream and 1.66 MB/s aggregate across 8 parallel range requests to one datanode from one IP (~145 GB/day) — against a local link calibrated at 27.6 MB/s, so this is server-side shaping, not a user. |
| Parsed numbers |
|
Evidence 9
| date | source | what it proved |
|---|---|---|
| 2026-10-21 | archive.org/developers/ias3.html | 'Fast GET downloads ... high performance downloads are achieved via the archive web infrastructure: curl --location http://archive.org/download/sam-s3-test-08/demo-intro-to-k.pdf' |
| 2026-10-21 | archive.org/developers/items.html | (snapshot 2025-11-08) — 'A particular file can always be downloaded from https://archive.org/download/<identifier>/<filename>' |
| 2026-10-21 | source not published | probe: curl -r 0-1048575 web.archive.org replay of a 406MB ISO -> HTTP 206, content-range honoured, accept-ranges: bytes, cache-control: max-age=1800 |
| 2026-10-21 | source not published | probe: 8 parallel 5MB range GETs to one replay target -> 40,000,000 bytes in 23s = ~1.7 MB/s aggregate; single 90s stream -> 1.19 MB/s; Cloudflare 50MB control download -> 27.6 MB/s |
| 2026-11-01 | probe: archive.org | an unauthenticated GET of an archive.org/download URL answered 200 after one redirect |
| 2026-11-01 | archive.org/developers/items.html | the download URL is a public URL, with no account named |
| 2026-11-01 | probe: archive.org | an unauthenticated GET of an archive.org/download URL answered 200, so the read path needs no identity |
| 2026-11-01 | archive.org/account/signup | an upload account asks a screen name, an email address and a password |
| 2026-11-01 | archive.org/developers/items.html | the download URL redirects to a datanode |
Unknown values 4
- per object cap notes: the field may be stated in a source that was not read
- per object cap stated: the field may be stated in a source that was not read
- retention notes: the field may be stated in a source that was not read
- retention stated: the field may be stated in a source that was not read