I found the vnstat project, which is already packaged for Fedora, and this is almost exactly what I needed.  After installing the package, I just needed to configure it:

1. Edit /etc/sysconfig/vnstat to make sure the correct network interface is monitored.
2. Edit /etc/cron.d/vnstat and uncomment the line that enables network interface monitoring (not sure why this comes commented out).
3. Edit /etc/vnstat.conf and check that everything looks right. I changed the locale, but I’m not sure why it doesn’t use the system locale.  Maybe something to do with cron?

It works by gathering data from the proc filesystem and using it to update a simple database file each time it is run from cron.

The only missing piece was the ability to examine network usage during peak hours only.  I’ve sent a patch for this to the maintainer, and put a small git repository here:

git clone http://cyberelk.net/tim/data/vnstat.git

If you’re reading this from the future and that link doesn’t work, hopefully it’s because the patch has been accepted…

I can now see my actual peak usage so far, as well as the estimated total for this month.  Here’s an example of the output (not real data):

 eth0  /  monthly

   month         rx      |      tx      |   total
-------------------------+--------------+--------------------------------------
  Jul '08       1.34 GB  |   103.00 MB  |     1.44 GB   %%%%%%%%%%%%%%%%%%%%::
   (peak)      32.57 MB  |     1.71 MB  |    34.28 MB
-------------------------+--------------+--------------------------------------
 estimated     16.10 GB  |     1.21 GB  |    17.31 GB
   (peak)        384 MB  |       12 MB  |      396 MB

Its purpose is to ask the user for a fax number when it receives a D-Bus signal from the hpfax backend. That’s all — so why does it sit there with a big fat ‘HP’ logo all the time? Unlike Windows, our notification area is not an advertising board.

Not that the logo is even in the notification area. Instead, it gets placed like a window, just anywhere there is space on your desktop. Nice.

That’s all disregarding the fact that none of this is any use at all if your fax queue resides on a networked CUPS server. It can send D-Bus signals if it likes, but they won’t be heard over the network.

The best one I came up with: ‘igh ‘eels.

Unfortunately it doesn’t seem to cope very well with having multiple ChangeLog entries added on the same date without separate date stamps, and I don’t know how to tell emacs to always add a new date stamp line when using it to add ChangeLog entries (C-x 4 a).

There is also support for proxy authentication. This is when you have submitted a print job to a queue for a network printer, but the remote printer requires authentication. When this happens, the CUPS backend responsible for transporting the job to the network printer stops and waits for authentication details. In system-config-printer 1.0.0, the print applet now spots this, displays an authentication dialog, and passes this to CUPS for it to complete the job.

This works for the ipp backend, but not yet for the smb backend. The samba v3-2-test branch has code to deal with this, but it doesn’t yet work for me (although I have a one line change that fixes it for me). As a result, the “prompt user for authentication” option just has the effect of not providing any authentication details at all at the moment.

Meanwhile, the job is held in the queue:

It resumes processing once the authentication details are supplied. Still to do: use GNOME Keyring to store the details, to prevent having to enter them over and over again when printing several jobs.

Here are some of the improvements:

• The configuration tool window has been made a lot simpler.  Instead of a list of printer names on the left and properties for the selected printer on the right, now there is just a box containing icons representing printers.  Double-clicking on a printer opens a properties dialog.
• The CUPS authentication dialog now selects the appropriate user-name, and allows it to be altered mid-operation.
• When the configuration tool is running, its list of printers is now updated dynamically.
• All jobs queued for a specific printer can be seen by right-clicking on a printer icon and selecting ‘View Print Queue’.  To see jobs queue on several printers, select the desired printers first before right-clicking.  To see all jobs, right-click with no printers selected.
• The job monitoring tool now displays a message when a job has failed.  If the printer has been stopped as a result, this is shown in the message.  A “Diagnose” button starts the trouble-shooter.
• The job monitoring tool now performs proxy authentication.  When a job is submitted but the CUPS backend requires further authentication information from the user, it stops the job.  In this situation, the job monitoring tool will now display an authentication dialog so that the job can proceed.

When 1.0.0 is released it will make its way into Fedora 10.  Depending on what sort of shape it is in, it might be released as an update for Fedora 9.

The printer properties and server settings are now in separate dialog boxes, although it doesn’t all work correctly yet.

I’ve checked this code in to the ‘new-ui’ branch.

Simply put, it goes like this:

>>> import smbc
>>> ctx = smbc.Context (debug=..., flags=..., auth_fn=...)
>>> dir = ctx.opendir ("smb://")
>>> print (dir.getdents ())
[<smbc.Dirent object "MYGROUP" (Workgroup) at 0x2aaaaaaf30f8>,
<smbc.Dirent object "WORKGROUP" (Workgroup) at 0x2aaaaaaf32d8>]


(Note that dir is an object representing a SMBCFILE* with a particular SMBCCTX*, and it calls closedir() automatically when it is cleaned up.)

It’s fairly easy to imagine it being extended to cover the rest of the libsmbclient API, but for now that is all.

I’ve put the tarball here.