Download websites that use gitweb

About gitweb

Nobody buys gitweb. The browser ships inside Git itself, so turning it on is a decision made by a systems administrator with a spare handler, not a procurement decision with an invoice behind it. What that means for a technology dataset is that the row is not a customer list. The row is an inventory of organisations that decided to host their own source code instead of renting the service from somebody else.

The tool is written in Perl and released under GPL v2, and the first release we record for it is 2005, close to the beginning of Git's own life. Age shows in the design: a script that renders repository history on request, with no daemon, no database and no queue behind it, which is exactly why it survives on machines that were set up long ago and never revisited.

The version data is the part worth reading slowly. We hold 3 distinct version strings for the tool, with 1.9 as the typical value. Because the browser is distributed with Git rather than separately, the version a host advertises is effectively the version of Git installed on that machine. Reading the spread of versions is therefore reading the age of the toolchain on each server, which is a more useful signal than the bare fact that a web interface exists.

There is a second reading, about what the choice excludes. A team running a packaged forge has committed to a product, an upgrade path and usually a database to back it up, while a host serving the browser that came with Git has committed to nothing beyond the version control system itself. The repositories sit on the same machine that publishes them, administered by hand. Finding the interface therefore identifies a working server rather than an appliance, and a habit of building infrastructure from parts rather than buying it assembled.

The people who want this list know why. Security researchers treat public repository browsers as an exposure question, because an interface meant for an internal network will happily serve commit history, branch names, author addresses and occasionally a configuration file that should never have been committed in the first place. Vendors selling hosted source control want the addresses of organisations still running their own. Anybody studying how infrastructure ages wants a population of long lived servers, and self hosted repository browsers are a dependable way to find them.

As of October 2026 we detect gitweb on 25 sites, and the figure deserves a caveat rather than a flourish. Detection depends on the interface being reachable, crawlable and recognisable from its markup, while a great many installations sit on internal networks, behind authentication, or on ports no scanner visits. What the dataset answers is which public hosts expose the interface, with the domains and versions attached. Anybody who needs the true installed base of a tool bundled with Git is asking something no external scan can settle.