The Andrew File System (AFS) was designed in the mid-80s at Carnegie Mellon University by a team with significant insight into how the computing ecosystem would operate in the future. The architecture they devised proved to be decades ahead of its time. In 1988, when the research project was completed Tim Burners-Lee's WorldWideWeb proposal had yet to be published. Corporations were not yet ready to provide Internet access to their employees or trust joining their networks (if they had them) to the Internet. The vast majority of Internet users relied upon dialup modem connections and almost no one had 24/7 access unless they were in a University dorm room.
In such a world, there was little public demand for a file system that:
* provided a globally accessible, location-independent name space
* used Kerberos network authentication
* provided a uniform ease-to-use Access Control List model for authorization
* used local caching to improve wide area network performance
* supported data replication and mobility for high availability
* access from a heterogeneous set of clients
It is not surprising that such a commercial product would find a small market. In hindsight, the best thing that IBM could have done with the intellectual property would have been to give it away. Thereby permitting AFS to supplant the need for HTTP.
Today, nearly 30 years after the start of the Andrew Research Project, the combination of features and platform availability have yet to be replicated by any other open or proprietary technology. While OpenAFS is considered complex. Its complexity stems entirely from the broad range of capabilities it provides.
I strongly recommend any organization that needs to share data among multiple offices or with a mobile workforce to consider deploying OpenAFS.