> Is the problem that sysisnt is running out of space on the root
> ramdisk?  if so, why not create (newfs) a new ramdisk and mount it as
> /tmp?  There's no need to carry `free [ramdisk[ space' around in the
> intialized data segment if we can create it on the fly.

That is of course an option, although sysinst should probably be
taught to (optionally?) do so.  However, this may cause more
problems in the _TINY case, as memory is pretty tight in that case
already, and you can't enable swapping (to expand the virtual
address space) before the disk has been partitioned.  Besides, I'm
not certain how much space sysinst requires on the ramdisk before
the disk has been partitioned.

I'm also not sure what we're going to do in the long run.  A few
points:

 o The install tools for 4M i386 machines and other i386 machines
   may eventually diverge more than they do now.

 o We may switch (?) to bzip in the boot code and elsewhere.

 o We may adopt the "ustarfs" method used on Alpha on the i386 to
   be able to load more than one diskette on boot (anyway, that's
   what I've understood is its primary motivation).

Regards,

- H=E5vard