Chuck Cranor said:
> 
> yeah, that is the exact conclusion i reached when i brought this 
> issue up to you and david greenman last spring when i was deciding 
> how best to handle this problem in uvm.   the issue is made more 
> complex by the fact that you need to track the references to parts
> of an object that may not currently be resident.
> 
The key is to remove the parts of an object that cannot be
re-referenced.  (There is where ref_count isn't the indicator
of being "re-referenced".)  For FreeBSD, we can handle swap
space issues almost as easily as memory (swap space manipulation
is virtual, and we don't have to swap-copy to collapse, when
needed.)  That is one reason for the 4yr old change in swap
addressing.

-- 
John                  | Never try to teach a pig to sing,
dyson@freebsd.org     | it just makes you look stupid,
jdyson@nc.com         | and it irritates the pig.