Fossil has self-hosted since 2007-07-21. As of this writing (2009-08-24) there are three publicly accessible repositories for the Fossil source code:
The canonical repository is (1). Repositories (2) and (3) automatically stay in synchronization with (1) via a cron job that invokes "fossil sync" at regular intervals.
Note that the two secondary repositories are more than just read-only mirrors. All three servers support full read/write capabilities. Changes (such as new tickets or wiki or check-ins) can be implemented on any of the three servers and those changes automatically propagate to the other two servers.
Server (1) runs as a CGI script on a Linode 1024 located in Dallas, TX - on the same virtual machine that hosts SQLite and over a dozen other smaller projects. This demonstrates that Fossil does not require much server power. Multiple fossil-based projects can easily be hosted on the same machine, even if that machine is itself one of several dozen virtual machines on single physical box. The CGI script that runs the canonical Fossil self-hosting repository is as follows:
#!/usr/bin/fossil repository: /fossil/fossil.fossil
Server (3) runs as a CGI script on a shared hosting account at Hurricane Electric in Fremont, CA. This server demonstrates the ability of Fossil to run on an economical shared-host web account with no privileges beyond port 80 HTTP access and CGI. It is not necessary to have a dedicated server to run Fossil. As far as we are aware, Fossil is the only full-featured configuration management system that can run in such a restricted environment. The CGI script that runs on the Hurricane Electric server is the same as the CGI script shown above, except that the pathnames are modified to suit the environment:
#!/home/hwaci/bin/fossil repository: /home/hwaci/fossil/fossil.fossil
Server (3) is synchronized with the canonical server (1) by running the following command via cron:
/home/hwaci/bin/fossil sync -R /home/hwaci/fossil/fossil.fossil
Server (2) is a Linode 512 located in Newark, NJ and set up just like the canonical server (1) with the addition of a cron job for synchronization as in server (3).