Lambda the Ultimate

Linear forwarders are actually the basic mechanism of an earlier implementation of the pi calculus called the fusion machine. We modify the fusion machine, replacing fusions by forwarders. The result is more robust in the presence of failures, and more fundamental.

And also:

The point of this paper is to solve the problem of input capability with a language that is “just right” – it neither disallows more features than necessary (as does the join calculus), nor adds more implementation work than is necessary (as does the fusion machine).

Yes, these are the same capabilities as in capability-based security. I am looking forward to read the complete paper, as it seems to confirm my unclear ideas of how capabilities and various pi calculi are related.