From my previous message > and David Laight ^ :

> Is there any way to install NetBSD so as to be able to track -current while
> not burning one's bridges of the stable distribution?

^Yes - everything is possible!

> Could this be done within one primary partition?
^Yes, doing it with 2 primary partitions is much more fun!
^You need the fixes to mbr_bootsel, the boot code and kernel autoconf code
^(see a load of bugs I've raised recently) to be able to boot 2 different
^copies of netBSD off different partitions on the same disk.

I guess then the primary partition for -current could be small, how big would it
need to be?  But I'd rather save primary partitions given the limited quota.

If I use LILO as the boot loader, there would be no problem finding the intended
partition (hopefully).  But I thought I read in earlier messages that NetBSD's
regular installer would not install to a second NetBSD primary partition,
however such a second installation could be performed manually.  I saw a
lilo.conf file, set up to boot Windows XP, FreeBSD and Linux, ironically, in
newsgroup comp.unix.bsd.openbsd.misc; it ought to work for NetBSD in place of
OpenBSD or FreeBSD.

> Maybe include relatively small f and g subpartitions for the
> experimental root and /usr?
^No need for a separate usr, bung it all in root

How big would this secondary root subpartition need to be, considering it would
be only for the base system, not the extra packages?

> Packages like emacs, mozilla, perl, vim 6, pine and others would be accessed
> in their normal installed places rather than being installed in duplicate.
^A few symlinks MIGHT sort that out.

Would these packages be part of -current?

> Is root invariably a?
^No - but you have to intercept the loader (while it is counting down)
^in order to select a different device. eg boot wd0f:
^Your kernel has to use the boot boot slice as its root fs as well.

I assumed root fs would have to be on the same subpartition as the kernel.