> [...] now I'm currently testing a much more efficient patch.  So far
> it looks like I can get 1.6 to run big long-running processes at
> least as reliably as I remember 1.3.2 running them, [...]

If anyone can give me shell access to an appropriate machine, I can try
the most effective test program I found (a mud - I was unable to make
it survive to the point of accepting incoming connections even once).

The machine would have to have at least some 32M of core, or the thing
will swap itself to death.  (The live process has a VM footprint of
about 26M right now.)  The kernel would also need whatever COMPAT_ foo
is necessary to run binaries linked static with a 2000-02-19 libc.
(Nothing fancy, no mount structs or uvmexp...but maybe stat(2), and
definitely sockets. :-)

Oh, it also has to be networked; I'm not going to copy 11+ MB of data
files over a console serial line. :-)

/~\ The ASCII				der Mouse
\ / Ribbon Campaign
 X  Against HTML	       mouse@rodents.montreal.qc.ca
/ \ Email!	     7D C8 61 52 5D E7 2D 39  4E F1 31 3E E8 B3 27 4B