Opal & Pilot

Mark Andrew SMITH (masmith@cs.ucsd.edu)
Mon, 17 Apr 2000 23:30:13 -0700 (PDT)


This paper looks at what assumptions we have made in the past due to a
small addressing capability of older machines. Now that addresses of 64
bits are showing up, what can we do differently to take advantage of this
and what can we stop doing that was an artifact of not having this
addressing capability in the past? The most obvious thing is the sharing
of data through pointers between processes becomes reality. Since both
programs run in the same address space, the resolution of the pointer will
be the same spot in physical memory, and the need to copy huge pieces of
data or to speak through pipes is eliminated. Protection is decoupled from
program execution, resource naming, resource ownership, and vitrual
storage. RPC becomes transparent through portals. This paper raises a lot
of interesting points and describes a good deal of tradeoffs, and mostly
benefits of a single address space.


An Operating System for a Personal Computer is presented in which the
authors examine possible security issues which can be overlooked on a
single user system. Security and IPC can be more lax because the single
user can be assumed to not be malicious. By ignoring some security issues,
we can get gains in performance. One interesting idea that may have first
been presented in this paper was the idea of memory mapping files. That
way, explicit I/O need not be used, simply by accessing memory, a file can
be modified or read. Some interesting ideas presented.

-Mark Smith