Later, Gnome adopted Sawfish, but you had a choice. I still used Enlightenment.

Then I don’t exactly recall what happened. I think Enlightenment went away from Gnome, I tried several (IceWM, Fluxbox, etc.) finally settled on FVWM, which I used for several years.

Then Liz made fun of me sufficiently for having an ugly WM that I switched to XFCE4.

XFCE broke their window stacking model, so I switched to Gnome + Metacity (the default).

This morning at work, Metacity decided to go insane, pegging my CPU and generally being unreasonable. Now, I had been annoyed by its sluggishness for quite awhile, so this was the final straw.

I installed lubuntu-desktop, and used LXDE + Openbox for today at work. It was a little.. interesting. Fast, for sure, but very minimalist, and not quite.. right in all cases.

On the drive home, it hit me… Gnome + Enlightenment.. again.

We’ll see how this goes. There is one little annoyance – with some themes, the gnome panel menu doesn’t seem to like non-minimized application windows managed by enlightenment… It makes you not able to click on menu selections.. However, with ShinyMetal (the theme I was originally using those many years ago), it works fine.

Oh, and I discovered Conky today, which may replace the gnome-panel flyout I have which does system monitoring. (I used to use GKrellM, but now I want trending for memory usage…

I will consider diaspora* when it is available.

If you want to talk to me, send me an email or comment on my blog.
If you want to keep me in the loop on your life, comment on a proper RSS-enabled blog forum to which I can subscribe.

As a result, I’ll be commenting here more frequently.

Edit: howto

I should note that it runs just fine under wine (tested: 1.1.31 aka the one which ships with Ubuntu 9.10).

Oh, and for the curious, the maintainers of the WineHQ AppDB drop “old” versions of Wine, without respect to what is currently shipping in distros. Try to be nice and submit a compatibility report and it will be rejected if you try to fudge the version number. KEEP OLD VERSIONS IN THE SELECT LIST. Lamers.

With additional notes from:
http://www.stgraber.org/2009/11/06/lxc-containers-or-extremely-fast-virtualization
http://linux-vserver.org/Upstart_issues
http://lxc.teegra.net/

1.) Note that you likely want to create a separate partition for machines and put the machines there.

2.) apt-get install lxc debootstrap
(Note – this is pretty inefficient in terms of space. The lxc setup does allow for read-only mounts of things. So, it can have its own /etc and html docs but share binaries and libraries with the host. But, I didn’t do it that way for ease of backups)

3.) Set up cgroup
mkdir /var/cgroup
mount -t cgroup cgroup /var/cgroup

probably want to add it to /etc/fstab as well:
cgroup /var/cgroup cgroup defaults 0 0

4.) Add a bridged network interface:
brctl addbr br0
brctl setfd br0 0
- bring up the bridge with a given IP
ifconfig br0 promisc up
- add the existing interface and set it to a bad
brctl addif br0 wlan0
ifconfig wlan0 0.0.0.0 up

5.) Bootstrap a basic system
mkdir machine
sudo debootstrap --arch i386 karmic machine

6.) Create a config file:
lxc.utsname = my_ssh_container
lxc.network.type = veth
lxc.network.flags = up
lxc.network.link = br0
lxc.network.ipv4 = 192.168.9.151/24
lxc.network.name = eth0
lxc.mount = /var/lxc-machines/machine/fstab
lxc.rootfs = /var/lxc-machines/machine/rootfs


7.) Create the machine:
lxc-create -n trac -f /var/lxc-machines/machine/config

8.) edit the appropriate files on the guest
/etc/fstab:
none /dev/shm tmpfs defaults 0 0
none /proc proc defaults 0 0
none /dev/pts devpts defaults 0 0
none /sys sysfs defaults 0 0
- /etc/resolv.conf
- /etc/hosts
- /etc/hostname
– Set them up correctly

/etc/init
- remove the following files:
rm control-alt-delete.conf hwclock.conf \
hwclock-save.conf mountall.conf mountall-net.conf \
mountall-reboot.conf mountall-shell.conf \
networking.conf procps.conf rsyslog-kmsg.conf \
tty2.conf tty3.conf tty4.conf tty5.conf \
tty6.conf upstart-udev-bridge.conf

- Create lxc.conf to generate things:
start on startup
script

       >/etc/mtab
       mount -a
       initctl emit virtual-filesystems
       initctl emit local-filesystems
       initctl emit remote-filesystems
       initctl emit filesystem
       ifdown eth0
       ifup eth0

end script

- /etc/apt
grab the sources.list from the host (if guest is same as host) or make one

- /etc/init/rc.sysinit.conf and change the line:
start on filesystem and net-device-up IFACE=lo
to:
start on filesystem
Important: This seems to be fragile in that system updates seem to put it back. If all your machines won’t boot suddenly, check this.

- bootstrap a basic bash so the various nice utilities work:
lxc-start -n trac bash

- add user.
- add admin group
- set sudoers to allow admin group
- add user to admin group
- Generate locale: sudo locale-gen en_US.UTF-8

9.) Start the machine:
lxc-start -n machine init &

Start for real:
lxc-start -n machine

10.) Notes:

Console:
sudo lxc-console -n machine

Debugging:
lxc-start -l DEBUG -o $(tty) --name machine

When running several:

java -jar something.jar

GNU ps actually shows which is which, because it includes the name of the jar file.

Windows just shows javaw.exe for all of them, so you don’t know what you’re killing.

Win.

(More to come. I’ll also be updating the community docs once I have all the kinks ironed out.)

Given:

• computers have become a sufficient commodity that everyone has their own
• disk space (eg fast local storage) is cheap

We can evaluate our objectives in that context:

1. Centralized application distribution is no longer necessary in that sense. Just add an additional in-house repository to your clients’ existing repository list and you’ve done the same thing, only with fast local storage. I mean, I suppose you could still install them to /usr/local, but it wouldn’t be a network mount.
2. On the individual machine side, I’d also tend to install things like video games in /usr/local, so that all users could share them – except there’s only one user on the machine, so what’s the point? For that matter, rather than installing whatever it was on two machines, just install it in your homedir and sync it amongst machines. For that matter, if it doesn’t work in the machine, just exclude it from the sync list.

I suppose that this isn’t really revolutionary, but I just realized that I was doing it stupidly, so I figured I’d share.

It isn’t helped by the fact that users are completely retarded and do things like stick MRI machines on the intertubes.

Since we make such systems at work, this was the subject of some lively discussion at work today, especially since we make a unit which connects home via a cellular modem.

This stuff can be really, really scary if not done carefully.

What is even more scary are old, insecure protocols, which were designed for isolated networks. Modbus, for example. Don’t put old bad stuff on teh intertubes!!!!