Cleanup of both imagezip and frisbee makefiles.

If any of these changes break something, please fix it and let me know.

"umount -A -t nfs".  As far as I can tell, the umount really works even with
the error.  I also cannot reproduce this manually.  Since we always intended
for a failure of this step to be non-fatal, I actually made it that way.

Why go after the root cause when you can treat the symptom!

- Tested to make sure it generates sane packets.
- Not yet tested on an actual emulab node.

packages' dependencies.  This is useful for creating a small Fedora Core
system without going through the installer.

Finally committed all of my Linux MFS stuff.  It's not done yet, but enough is there that I need to get it into version control before I lose it.

This includes:
- Build tree for a Linux frisbee MFS (the source tarballs are in ops:/share/linux/mfs/source)
- Patch for pxelinux to add bootinfo support
- Patch for grub to support our magic config sector

be in elab mode somehow boots from the PLC by accident...

discover resources call.

"modes".  Also some machinations in libplab and the federation to support
not only object synch, but also "operations".

Added a few more OS support functions to liblocsetup.
Modified ifcfg-eth99 to behave on FC9.

username out of the db.  Don't want to try slogin as root from a script.

Added scripts to support a Linux MFS.  Also began modifications to rc.frisbee and slicefix to make them less FreeBSD-specific.

Lots of corner cases that I am ignoring and recovery from errors is
just as crude. See xmlrpc/client.py ...

uid of the user. Not sure this change will survive in its present form
since this was done for protogeni and the spec will probably call for
something different. But this is what it is for now. Note that the server
side is backwards compatable with old certs, at least for now.

evproxy.

case). Still more to add.