Around 20 years ago, enthralled by the database-like properties of the Be File System in BeOS, I decided to see if I could use the filesystem itself as a backing database for a live web server application. I ultimately wrote TrackerBase, and served betips.net from a BeBox in my home for a few years.
The basic idea was that BeOS let you create custom filetypes with custom attributes, and provided APIs for querying those attributes. The rest was just custom database query code, but without an actual database.
Fast forward two decades. I’m all-Mac now. I recently stumbled across the original .pkg installer for TrackerBase and wanted to have a look. Discovered that there was no way to open the package from the Mac side, and reached out to the Haiku community (Haiku is the open-source successor to BeOS). Within an hour, a community member provided me with a tarball, and I was able to see that code again for the first time in a long time.
It’s always embarrassing to look back on old code, but that was especially true in this case. And Perl? What was I thinking? Anyway, these days we have github, so I’ve just published the original code on github, untouched, for posterity’s sake. MIT license.
Irrelevant as it may be today, I’m still proud of it as a novel application of an innovative filesystem.