Download counts#

This page describes what the download counts shown on Sysand Index mean and how they are computed.

What is counted#

A release download is counted when a client fetches the release archive (project.kpar) for a specific version from the index. Each version keeps its own count; a project’s total is the sum across its versions.

The following are not counted:

  • HEAD requests for the archive (used to check availability or size).

  • Fetches of release metadata (versions.json, .project.json, .meta.json).

  • Re-fetching the archive directly from the object storage URL the index redirects to, without going through the index endpoint again.

Deduplication#

Downloads are deduplicated per downloader: repeated downloads of the same release from the same client address within a 24-hour window count once. A count therefore approximates unique downloaders per day rather than raw requests.

Deduplication is best-effort. A download may occasionally be counted more than once — for example around server restarts — but real downloads are never missed.

Where counts appear#

  • Per-version counts on a project’s versions page.

  • Project totals on project pages, in search results, and in the “Most downloaded” list on the home page.

  • The downloads sort option in search.

  • The site-wide downloads total on the home page.

What counts do not measure#

Download counts are an activity signal, not an install or user count:

  • Machines sharing one address — such as CI runners or hosts behind the same corporate network — count once per day together.

  • Local caches mean a project in active use may be downloaded rarely.

  • A download says nothing about whether the project was kept or used.