In article <20000821134557.B27866@cs.swarthmore.edu> you write:
>Since the vnode pager (at least, assuming the comments in
>uvm/uvm_vnode.{h,c} do not lie) swaps to vnodes, which are handled by
>the vnd device driver, which must reside on the local file system,
>it sure seems to me that vnd(4) *is* involved and that they are not
>useful for what I am trying to do.

Where do you get the impression that vnd(4) is connected to the vnode pager? 
As I understand it, vnd(4) is a means of making the contents of a file
available through a block-device interface, while the vnode page makes the
contents of a file available through the virtual memory system.  As far as I
know, they're independent of one another.  It's certainly possible to build
a kernel that uses the vnode pager (it's not currently disablable) but
without vnd(4).

>I'd be thrilled if someone criticizing my approach would like to do so
>constructively.

If nothing else, discussions of the VM system belong on tech-kern, where
people who understand it (unlike me) hang around.

-- 
Ben Harris                                                   <bjh21@netbsd.org>
Portmaster, NetBSD/arm26               <URL:http://www.netbsd.org/Ports/arm26/>