After running funkload tests [1] on CentOS 5.7 I ran into a problem: I was not able to generate the reports.

Digging into the funkload internals, I realized that the problem was related to a syntax error raised by gnuplot, that was silently swallowed by the report generator command.

The funkload generated plot commands were in conflict with the gnuplot version installed by default on the system (4.0 patchlevel 0, a very old one!).

On my shiny kubuntu I had no problems with gnuplot version 4.4.4, so I solved this problem by adding an updated gnuplot version to my buildout.

Here is how I made it work:

[buildout]
extends = base.cfg
parts+= 
    gnuplot
   
[gnuplot]
recipe = zc.recipe.cmmi
url = http://sourceforge.net/projects/gnuplot/files/gnuplot/4.4.4/gnuplot-4.4.4.tar.gz/download 
configure-options =
    --bindir=${buildout:directory}/bin

For this trick to work properly, the bin directory inside the buildout folder should be in your PATH environment variable. If you are using virtualenv properly, you will probably stumble upon this.

[1] Version 1.11.0, the one that, to my knowledge, works best with collective.recipe.funkload as of version 0.3.1

In my daily activity I find myself connecting to a lot of servers and it is hard to remember for each of them the correct hostname or IP and the related credentials.

To speed up my work I try to keep has much as possible updated my .ssh/config file.

Let's say, for example that one of our clients, let's say ACME :), gives us access to a fictitious server with the IP address 999.999.999.999. 

They give us the access that server with the username bugs and ssh is listen to port 2222.

As soon as I am authorized to connect to a new server I insert a new section to the file like this:

# The new acme server
Host acme
HostName 999.999.999.999
Port 2222
User bugs

 This way instead of typing:

ssh -p2222 bugs@999.999.999.999

I can just type:

ssh acme

with the same effects.

The benefits of that go beyond the use of ssh, because the same configuration file has effect on other ssh related programs like scp, rsync and sshfs.

In addition .ssh/config is parsed by bash completion scripts so I can save further keystrokes (holy laziness)!

This is good, more than for today, for tomorrow because time passes and I can forget the username, the server IP or the port, but I would hardly forget I worked for ACME and to get those info quickly reading .ssh/config is enough.

Of course I can add many hosts sections to the .ssh/config like I am showing in this example:

## List of ACME servers ##
# The staging server
Host acme-staging
HostName staging.example.com
User bugs
# The production server
Host acme-production
HostName production.example.com
User bunny

## Another customer ##
Host demo
HostName 10.0.1.100
User donald

# I can also setup a tunnel to a port
Host demo-debug
HostName 10.0.1.100
User root
LocalForward 8080 localhost:8080

As you can see each section starts with the Host declaration and the settings are effective only for that specific Host.

In the last example we make a tunnel forwarding the 8080 port to localhost, useful in case the server firewall filters direct connections from outside.

With this configuration the command:

ssh demo-debug

is equivalent to:

ssh root@10.0.1.100 -L 8080:localhost:8080

Further readings

Beyond the nice manual page about ssh configuration:

man ssh_config

you could find very useful those links to

• SSH authentication using keys (passwordless)
• SSH Agent Forwarding
  

A note to KDE users

.

It is the second time in a couple of days that I am asked why in some pages there is no link to the @@manage-portlets view. In both cases the archetypes missing the link (and the view) where inheriting from a very basic object.

After a small research I found that this view is registered for the interface plone.portlets.interfaces.ILocalPortletAssignable.

Making the object interface inherit from this interface solved the problem.