The key question, of course, is ``What kind of lunatic builds a new operating system these days?''

There are a number of reasons to do so:

• From a purely research perspective, it would be interesting to know if a fast capability system is possible. KeyKOS came very close, but EROS is looking to go a step further.

• Also, we're interested in building systems from components, which requires access delegation. The idea here is that you want to be able to load up some object and give it access to a file and a window to display in without giving it access to your entire user environment.

• Separate from all of this, my feeling is that persistence simplifies a lot of applications, including things like databases. With the benefit of hindsight it was a mistake to conflate the capability performance experiment with the persistence experiment, but I think we'll see shortly that in the end EROS doesn't come out too badly.