Is Systems Software Research Irrelevant?

["Anton van Straaten" <address@hidden>]

This may be well-known, but I happened across a wonderfully pessimistic
document/slideshow during a Google search: "Systems Software Research is
Irrelevant", by Rob Pike of Bell Labs, from Feb 2000:
http://www.cs.bell-labs.com/who/rob/utah2000.ps
http://www.cs.bell-labs.com/who/rob/utah2000.pdf

Regardless of whether one agrees or disagrees with the overall premise, I
thought it contained some good points, many of which are quite applicable to
the intersection of academic vs. practical issues discussed on this list.

One of the claims is that due to the scale and long timelines involved in
large projects like operating systems, academia and research groups in this
area are marginalized, forced to restrict themselves to measurement,
microspecialization, and tweaking.

Although the focus of these slides is primarily on traditional operating
systems research, it seems to me that the situation with large virtual
machine environments like JVM and .NET may be similar (they aren't that far
from operating systems, anyway).  Research groups presumably can't match the
kind of effort that goes into the development of these systems (?), so may
be similarly constrained in the problems they can address while remaining
relevant.

Of course, talking of "relevance" brings to mind Dennis Ritchie's statement,
"The greatest danger to good computer science research today may be
excessive relevance".  This document is about both the the difficulties of
being relevant, and the need to be relevant.

Don't read the document if you are prone to fits of depression about your
work!  ;)

Anton