IO error in swap. Solution?

[Tim Pointing]

markb at sgi.com (Mark Bradley) says:
> IO error in swap is fatal to the machine, as unix has no way to know what
> to do about missing memory.  This is not unique to any [SGI] release.

Why does a swap error cause a panic (other that saying that that is the way
SysV does it)? Bekeley-derived versions of unix simply kill the process which
got hit by the swap error ("process <PID> killed due to I/O error in swap").
This does, in most cases, allow for a graceful, controlled shutdown of the
system.  Is there something in the SVID spec that says that the system has
to die on swap errors? Is this something that could "easily" be added to the
next SGI OS release (3.7?)?

--
Tim Pointing, DCIEM
tim at zorac.dciem.dnd.ca / tim at zorac.arpa