no freebsd support.

The primary change is that tmcd now sends down a list of setting to
apply to each interface, and that list is turned into a hash table,
and provided to rc.config, which passes them along to the machine
dependent routine in liblocsetup.

Then in the linux version of liblocsetup there is a bunch of new code
to configure wireless links using iwconfig and iwpriv, using the
settings array.

All of this is bound to change once we get more experience with it.

mentions you need to restart the event system to fire off the start
command again.

Currently, you cannot name the startcmd in a tevc command cause its
name is unknown. Something to do with how objects are renamed in the
parser and that they are not done below top level (inside a
subroutine). Not sure how you get around that.

there is not default imageid, so it can generate an error!

arbitrary interface settings associated with the virtual part of the
topology.

Declaring the host is not required.
Point to internals slides.

where we do other interface table cleanup).

* Return interface_settings info for the node, across all interfaces
  for which there is info in the interface_settings table.

* Prepend tags; INTERFACE and INTERFACE_SETTINGS to make it possible
  to tell them apart at the client.

intuit the default value from BOSSNODE .

settings to the nodes when they configure their interfaces. Added primarily
for wireless interface support.

not to run linktest.

invoked from the RPC server.

down when invoked from the RPC server.

probably not the final form for this stuff, but is fine for initial
development of wireless support.

	tb-set-protocol $lan0 "80211g"
	tb-set-lan-accesspoint $lan0 $nodew1

an "is_accesspoint" boolean to virt_lans. The former defaults to
"ethernet" but can be set to anything (80211a, 80211b, etc) in the NS
file. The is_accesspoint is temporary, and simply allows you to set
which node is the accesspoint in the NS file. This slot will probably
move to another table at some point.

Added interface_capabilities table, which is intended to list the
capabilities and the default values, for interfaces listed in the
interface_types table. This allows a more flexible description of
interfaces, expecially wireless devices. Initially, I have seeded the
table with just the default protocol (ethernet) and the speed. For
example, the fxp:

	fxp              | protocols         | ethernet |
	fxp              | ethernet_defspeed | 100000   |

As you can see, protocols is plural, and is intended to be a comma
separated list. So, for the atheros wireless card:

	ath              | protocols         | 80211a,80211b,80211g |
	ath              | 80211a_defspeed   | 54000                |
	ath              | 80211b_defspeed   | 11000                |
	ath              | 80211g_defspeed   | 54000                |

I gave up on using the entire row as a primary key; this is just too
painful from perl/php/python where hashes are the most popular data
structure, and duplicate columns get overwritten.

with every DOS partition.  It was only used by NTFS and has been replaced
by equally hacky changes to the NTFS library.

The benefit of the latter is that it at least works all the time on
Linux and all FreeBSD versions.

Add a new interface to the library to enable us to use the whole-disk
device rather than having to (figure out and) use a specific slice device.

statement.

to build under FreeBSD 5 (really newer gcc's)

Install emulab-sync man pages.

Install emulab-sync man pages.

Usability improvements to the emulab-sync programs:

  * Terminating a TCP client before the barrier is reached will back
    out any changes to the barrier's count.

  * You can SIGHUP the server to clear all of the barriers and release
    the clients with an exit value of 240.

  * Basic error reporting is supported using the "-e" option and the
    program's exit code.

  * On FreeBSD, the server will respond to a SIGINFO (^T) by printing
    out the list of active barriers.

  * More checking of the inputs.

Still, will need to watch for temp file buildup.

site.  I changed the headers on the parameter listings to use <th>
instead of <td> since they look too much alike otherwise.