These need to be generalized since its probably not the last non Emulab
aggregate we are going to talk to.

the source is the geni-lib script.

what they should be from the siteids on the nodes.

I thought I would be clever and check the client ID on a JOIN reply
to make sure we match up the proper request/reply (an issue when replies
are multicast). However, on a JOINv1, the reply does not include the
clientid. Unfortunately, for previous backward-compat situations, we
still do generate JOINv1 requests most of the time.

So for JOINv1's we don't try to check the clientid. This means that
with a newer server that MCs replies, a JOINv1 client may see someone
else's reply and start transmitting requests even before the server
has seen that client's join request. This is not an issue for us right
now since the server only loosely tracks its clients and does so just
so that it can report stats when the client leaves. I say "loosely"
because, e.g., the server does not require that block requests come
from a JOIN'ed client.

Previously on MFMH:

 * Our hero added incremental kernel export updates to mountd so
   that a single update would finish within his lifetime,

 * Hero subsequently discovers that an incremental update still
   takes 8 seconds on the Emulab Mothership,

 * Hero discovers that remaining time is spent NOT in kernel operations
   or even complex calculations of orbital mechanics, but rather in simple
   parsing of text from a 1300 line file.

 * Hero gives up in despair.

This week on MFMH:

 * Our revitalized hero applies gprof-fu to the problem and
   discovers that 7.9 seconds of the 8 second /etc/exports parse time
   is spent stat()ing /etc/nsswitch.conf 2,392,057 times.

 * "WtF!", exclaims our hero, "that is 27.3 stats for every text
   character in the file!"

 * Problem is traced to the fact that we do "-maproot=root" for each
   of the 1344 exports and every parse of that option looks up root
   in the password file and determines which of the 1784 groups in
   /etc/group root belongs to. Each of those 1344 * 1784 "get group
   entry" calls, stats /etc/nsswitch.conf to see if the file has
   changed.

 * "Bite me!", sez Mike, and he adds a CACHE_ROOT_CREDS option to
   mountd so that the lookup is only done on the very first instance
   of -maproot=root.

A single incremental update of exports now takes 0.75 seconds.

June 15 (noon).

It used to mean bytes per second.  Now it means bits per second.  To
avoid worrying about client-side versions and updating, we'll keep the
client unmodified and convert statistics on their way through the server.
Who doesn't like ugly hacks?

certificate. Need to decide how to deal with it; the cert parsing code
requires a URN in the certificate, and I don't want to mess with that right
now.

handling.

with tag fragmentation which can causes us to not find any free tags
even there are plenty. Add some debugging when we do fail to find
a free tag.