Stephen Kell's research prehistory

Original PhD proposal

In May 2005, shortly before completing my undergraduate studies, I drafted a proposal entitled "Thread Tunnelling in Vertically Structured Systems". It's only really of historical interest now, but you may read it if you like.

Undergraduate dissertation

Although it's not exactly research, during the final year of my degree I completed an individual project entitled "A General-Purpose Synthetic Filesystem". The project investigated the possibility of using the Unix filesystem as a universal interface for application scripting and debugging (and perhaps also for more general instances of IPC). An arbitrary set of Java objects can be exposed as a filesystem with essentially a one-time coding effort, thanks to the the run-time type information of the JVM. This contrasts with filesystems such as Linux's /sys, which do not have access to run-time type information, and so typically require new code for each type of object exposed.

I implemented a filesystem exposing a graph of Java objects, using the fuse user-space filesystem package for Linux. Unfortunately, the resulting system was rather inefficient and difficult to evaluate. Nevertheless, my dissertation provides a flawed but interesting discussion which highlights some of the limitations of conventional filesystem APIs. You can read it here. If you'd like to play with my code, just contact me.

Content updated at Tue Jul 18 19:02:14 CEST 2006.
validate this page