Category Archives: Technical

Google Drive and Linux

Google Drive is a handy way to share files across multiple computers, or back up files into the cloud. Naturally it integrates seamlessly with Android, and Google provide a client for use on Windows and the Mac, so you can keep you tablet, phone and PC all synced up. Unless you use Linux that is. Because unlike Dropbox, there is no official Linux client for Google Drive. Fortunately, help is at hand in the form of Insync.

Insync is a full featured proprietary Google Drive client. I’m focusing on Linux because it was trying to find a Google Drive client for Linux that led me to Insync, but it is also available for Windows and Mac. Unlike the official client, it supports multiple Google accounts, if you have a need for that sort of thing.

I have it installed on a laptop running Ubuntu 12.04 and it seems to do what it says on the tin. It constantly monitors a configured directory and keeps it in sync with my Google Drive account. It can handle having symlinks in the directory, so for example I’ve symlinked my Shotwell data directory to make sure that is backed up. And it can automatically convert Google format documents to LibreOffice format on the fly. It also supports selective syncing, allowing you to exclude certain folders from syncing. I’ve only had it installed for ten days or so, so it’s a bit early to tell if it functions reliably, but I’ve had few problems so far. The one niggling issue is that sometimes it fails to start at system startup, but support from Insync staff has been responsive so I suspect this will get sorted out soon.

Much Linux software is open source and distributed free of charge, so using proprietary, paid for software may come as a bit of a shock to some Linux users. If you have some deeply held philosophical objection to closed source this is not going to be for you, but otherwise it is definitely worth a look. The cost is far from prohibitive—the consumer version is $10 per Google account. When you consider that beyond the basic free allowance Google Drive is significantly cheaper than Dropbox you could quickly save this if you need more than a few gig of space. Cost over a year with 100GB of storage, including software purchase, would be $69.88 with Google + Insync consumer[1], or $99 with Dropbox[2].

If you want to try it out Insync is available on a 15 day free trial. To be honest, unless you make a lot of use of Google Drive this is not really long enough to evaluate something that just works away in the background. It’s also perhaps one place where the Insync team are missing a trick—it’s certainly not long enough to get you hooked. I suspect if they bumped this to six months they’d find a lot of people install it, forget about it, learn to take it for granted, and then pay for it a soon as they realise they’re about to be without it.

  1. Based on $4.99 per month + $10 for Insync. Google Drive price from, retrieved 24 Oct 2013 []
  2. based on $99 per year. Price from, retrieved 24 Oct 2013 []

Call popups with Asterisk and Thunderbird

There are several applications around for Windows users that offer popup notification of incoming calls to an Asterisk server, looking up the caller’s name in various Personal Information Managers (PIMs). Since I don’t use Windows, they’re not much use to me, so I wondered if there was anything available for Linux.

A Google a while back threw up this script by Olivier H. Beauchesne which looked promising. It does part of what I wanted—it generates a popup notification of incoming calls, but it makes no attempt to look up the name of the caller.

I considered writing something as a Thunderbird extension, but there seemed to be a couple of downsides to this approach:

  • From recent experience of working with Thunderbird’s address book for a click2dial extension I knew that it doesn’t lend itself well to look ups keyed on phone numbers
  • I wanted something that would always be running, irrespective of whether or not I’d fired up Thunderbird

I decided to have a go at hacking Olivier’s script, quite ambitious considering I’d never written anything in Python before.

It didn’t actually take much to get it to look up incoming numbers in a SQLite database. I then wrote a dirty Perl script to pull all the phone numbers out of my Thunderbird address book and populate the database. The Perl script was way to ugly to share and I always intended to tidy it up, but just never found the time. Until now.

callPopPy in action

callPopPy in action

A prolonged period off work sick has driven me to looking for things to keep my mind occupied, and so I’ve returned to this project. The Perl script has been scrapped. In it’s place I’ve written a Thunderbird extension named Squalit, a much neater solution. Squalit can export a single contact or an entire address book to the database, and can be configured to update the database periodically, ensuring that call popups and Thunderbird are always in sync.

I then turned my attention to the popup script itself. The original relied on libnotify and its Pyhon binding, pynotify. These are only available on Gnome based Linux distributions such as Ubuntu. I thought it would be useful to make everything more portable, so rewrote it to use a library I stumbled across by Daniel Woodhouse, gtkPopupNotify. After a fair few other changes I was left with a distant descendant of Olivier’s original script, which I’ve called callPopPy. callPopPy is portable enough that I’ve had it running on a Windows XP machine, and is doing just what I wanted on Ubuntu. The advantage of the two stage approach is that other utilities could be written to integrate callPopPy with other PIMs.

The combination of TBDialOut for click2dial with Squalit and callPopPy for call popup notification provides me with great integration between Thunderbird and my Asterisk server—I guess this is what is meant by CTI.


Dialing from Thunderbird’s address book: TBDialOut

I’ve had my first go at developing an extension for Thunderbird, the open source email client from Mozilla. The unimaginatively named TBDialOut is an extension that enables the user to place a phone call direct from Thunderbird’s address book.

I’ve actually been using an extension that does this for quite a while—VOIP3Dialer 1.1. But I recently upgraded to Thunderbird 3.1.7 from 2.something-or-other and VOIP3Dialer didn’t work with TB3. I really wanted to keep the funcionality that VOIP3Dialer offered, and with it not having been updated since July 2008, and its support web page gone it seemed there wasn’t much option but to learn how these extensions things work.

It didn’t take long to work out what had changed in Thunderbird’s address book to break things. That sorted, I started to look around a bit more and get to understand the inner workings of an extension. I ended up rewriting pretty much all of the functional code in VOIP3Dialer, and then started adding stuff.

TBDialOut introduces various user options, including a choice of URL schema to use and international dial prefixes. I then tackled the user interface. I added clickable links as well as the original context menu items and three buttons and made a few other usability tweaks. Finally, I changed the three buttons for a single, combined button that prompts the user to select which phone number to dial (home, work or mobile (cell phone)).

Screenshot of TBDialOut

TBDialOut in use

TBDialOut basically just passes the selected phone number to which ever programme is configured to handle the URL scheme you’ve chosen to use with TBDialOut. You can choose from sip:, tel:, callto: or skype:. I’m using it with a small Perl script that contacts my Asterisk server to set up a phone call. When I click on a number with TBDialOut the phones in the house ring out. Pick one up and you’re connected to the number you clicked.


Pifimon: NetStumbler for Linux?

Well no, not really. But Pifimon does do a little of what NetStumbler does.

Pifimon is a small programme to monitor wireless networks under Linux. It is written in Perl and works by presenting the output of iwlist in a more friendly, and constantly updating, way.

The initial screen presents a list of visible access point with a summary of information about them in table format. The list is updated as fast as iwlist can rescan. You can select one access point to monitor and display that access point’s signal strength as a constantly updating histogram. Here are some screenshots:

List of access points

Initial screen: list of visible access points

Signal strength histogram

Monitoring the signal strength of one access point

Pifimon supports several wireless card drivers, and provides a way to extend support to other drivers. You can download the latest version (pifimon-0.4rc2.tar.gz) or get Pifimon from Github.

Why did I write it?

For various reasons I’ve become interested in building antennae to boost wifi coverage. I needed some way to get a constantly updating representation of signal strength to help test and align antennae. The thing most people seem to use is NetStumbler, but NetStumbler only runs on Windows. Most of my machines, and in particular the netbook I plan to use for testing, run Ubuntu.

After a bit of Googling I came across scanmeter, a bash script that processes the output of the Linux `iwlist scan` command to produce a histogram. Just what I needed! Well nearly, but the great thing about open source is that if it’s not quite right, you can change it.

I started playing about with scanmeter. I didn’t want to have to enter a whole MAC ID to select the cell I wanted to monitor. I thought it would be nice to colour code with a third colour for ‘moderate’. I put colours in the signal strength column of the cell list. Then I thought it would be nice if the cell list updated regularly…

By now I was really stretching the limits of my bash scripting. So I decided to rewrite it in Perl. A bit more googling threw up this blog on using Perl to neaten up iwlist’s output which provided some great ideas on parsing iwlist output with Perl. I rewrote that script as a Perl package, largely to allow for easy extending to cope with different wireless drivers producing different iwlist output. A few hours later I had my first stab at Pifimon…



Music to my ears

It all started when I realised I could connect my netbook to my hifi to play music, and then sit back on the sofa and control the netbook from my laptop using VNC. Wonderful! I’m lazy by nature and changing the CD is such a drag. I found I listened to so much more music this way, and really enjoyed it. So I began looking for a more elegant, and more importantly, better sounding solution. This is what I came up with…


This is going to be a very techy article, so if that’s not your thing, turn away now.

I’m using an Asus EeePC 701 that I picked up on ebay for £71 to play the music. It basically has Ubuntu installed, but tweaked to be as minimal as possible. Music is on the file server upstairs, mostly as FLAC files. The EeePC runs Music Player Daemon (MPD), a music server that can be remotely controlled across the network by a variety of clients. A very minimal GUI is loaded at startup with the Gnome Music Player Client (GMPC). Starting everything is as simple as pressing the power button—no logging in or starting things up. Also on the EeePC is Remuco, which allows remote control from my Nokia phone.

Rather than relying on the ropey internal sound card, the digital signal is taken from the USB port to a Cambridge Audio DacMagic upsampling DAC, and on to my existing Cambridge Audio Azur 340A amp.

The EeePC

EeePC with Gnome Music PlayerI chose the Asus EeePC 701 for a number of reasons. I have an EeePC 901 and I’ve been very pleased with it, so I trusted the pedigree. It has a solid state hard drive, so there is no clattering drive noise to interfere with listening. It’s small and fits easily on the shelf next to the hifi. And second hand ones come up regularly on ebay at a reasonable price.

It might have been a bit of mistake to be honest. I hit a couple of problems. One is that it seems to run hotter than the 901, so the fan comes in more, but to be fair it’s not really noticeable. More of an issue has been the wireless chipset – the Realtek RTL8187SE. Support for this under linux seems incomplete, and though it works fine with NetworkManager once a user is logged in, I couldn’t persuade it to come up at boot time. Consequently, I ended up investing in a Zyxel G-202 USB wireless adapter instead. I did manage to kludge the internal thing to work using NetworkManager, but that meant the network wasn’t started until late in the boot process, which delayed various other things. Switching to the well supported Zyxel adapter shaved 20s off the boot time, taking it down from 65s to 45s.

The DAC Magic

Cambridge Audio DAC MagicThe Cambridge Audio DAC Magic is a lovely piece of kit. It’s an upsampling DAC with more acronyms in it’s specification than you can shake a stick at. Even I’m not geeky enough to understand what half of them mean. But it sounds great, and what else matters? Well, the fact that it does what I need matters actually, which is to take USB input from a PC. And the fact that it seems to be fairly easy to set up a linux based machine to recognise it as a sound card.

Getting started

My starting point was to install Ubuntu Netbook Remix on the EeePC. A slimmer distro might have made sense, but I figured UNR has been tweaked for various netbooks and I might save a fair bit of messing about with hardware drivers if I started with it.

Having installed UNR I created a second user named ‘music’, as an ordinary desktop user. This will be the user who will automatically log in to the EeePC. I’ll keep my own account for administration.

I then used Synaptic Package manager to install mpd, gmpc, gmpc-plugins and remuco-mpd—the basis of the music system. I needed a couple of other packages too. nfs-common enables me to access the music stored on the fileserver upstairs; and mingetty will help me to automate some login tasks.

The aim was to have a system that can be turned on with nothing more than a touch of a button, and that boots fairly quickly, bringing up as little of the GUI as possible. Various suggestions were made to me about session files and using gdm, but I found it actually quite difficult to use gdm and not get a lot of extra desktop guff. So I gave up.

Instead, I disabled gdm altogether by moving its upstart file:

# mv /etc/init/gdm.conf /etc/init/gdm.conf.disabled

Now the machine will boot to a console login prompt. The next job is to get it to login. This is where mingetty comes in. Open /etc/init/tty1.conf and edit the ‘exec’ line to read

exec /sbin/mingetty --autologin music --noclear tty1

Now you should have a machine that will automatically login the user ‘music’. You can still login as another user by hitting Alt+F2 to bring up an alternative console (tty2). You can then start the full desktop GUI by typing

# startx

Bringing up the network

The simple way is to configure your wireless network in NetworkManager and then check the “Available to all users” box. NetworkManager won’t start a wireless network until a user logs in it seems, but a console login will do the trick. The trouble with this approach is it delays the start of the network, and I found that in order for things to work I had to be sure the network was up and the remote directory with all the music in mounted before starting MPD. So I looked for a way to start the network earlier.

As I mentioned, this proved problematic with the built in wireless card, and I had to resort to a USB wireless adapter. Having used ifconfig to establish that this was seen as wlan2, I edited /etc/network/interfaces

# The loopback network interface
auto lo
iface lo inet loopback

# The primary network interface
auto eth1
#iface eth1 inet dhcp
auto wlan2
iface wlan2 inet dhcp
wireless-essid <yourssid>
wireless-key <yourhexkey>
wireless-mode managed
wireless-channel <yourchannel>

and bingo! The network comes up early in the boot process. Of course, this assumes WEP security on the wireless network. If you need WPA or something else you’ll have to search further…

The network share

Connecting to the network share should be as simple as putting the appropriate entry in /etc/fstab. Unfortunately, it wasn’t.

The problem seems to be that /etc/fstab is processed before the wireless network has finished coming up. It turned out to be fairly critical that things are done in the right order, and that one thing finishes before the next starts. Otherwise odd things happen.

  • Start the network
  • Mount the network share
  • Start MPD
  • Start remuco-mpd

So the entry in /etc/fstab is flagged ‘noauto’ and ‘user’. It is not mounted automatically at boot, but by a script (we’ll come to that) that runs once the network is up. In my case my fileserver is at and it exports /home/shared via nfs. I’m going to mount this at /network/shared on the EeePC, so I have this in /etc/fstab    /network/shared  nfs   noauto,defaults,user,ro,bg   0  0

Setting up MPD

Configuration of MPD is done mainly by editing /etc/mpd.conf. The distribution includes sensible defaults for most things, and plenty of comments. You’ll need to set music_directory to the directory beneath which all your music is stored. Also check the value of bind_to_address. Do not set this to ‘localhost’ if you want to able to control the system remotely. I simply commented this line out, as the default is to bind to all available addresses.

Configuring MPD to output to the Cambridge Audio DAC magic proved to be a doddle. On plugging in the USB cable, the DAC magic was recognised by ALSA and assigned as card 1 (the onboard card gets slot 0). You can check which card has which slot with

aplay -l

All it takes to get MPD to output to the DAC Magic is this in the audio output section of /etc/mpd.conf:

audio_output {
 type        "alsa"
 name        "DAC Magic"
 device        "hw:1,0"
 mixer_device    "default"    # optional
 mixer_control    "PCM"        # optional
 mixer_index    "0"        # optional

The next thing to do is to disable MPD. I know, that sounds weird. It’s back to this problem of the order that things are done. By default MPD is configured to start automatically at boot time. The problem is, it gets rather unhappy if it is started when music_directory isn’t there yet. So we’ll stop it from starting at boot, and start it later from a script.

sudo update-rc.d mpd disable

Starting things up

Now it’s time to consider how to start everything up in the right order. This is the ugliest bit of the whole set up. Because my music is on a fileserver upstairs in the office I need the network to be up before I can really do much else. Fortunately, Ubuntu provides a way to run scripts in response to the network coming up—simply place a script in /etc/network/if-up.d

This is the script I used. It is quite specific to my set up, so you’ll need to go through and change stuff. Basically, it runs only when the wireless network comes up and then uses a series of while loops to wait until things are ready before moving to the next stage.


# if the interface coming up isn't our wireless card,
# bail out (otherwise the script runs when the loopback
# interface comes up, which isn't useful!
if [ "$IFACE" != "wlan2" ];
 exit 0

# wait until we're sure the network is up
NET=$( ( ifconfig | grep 192.168.37 ) 2>/dev/null )
echo -n "Waiting for network ."
while [ "x$NET" = "x" ];
 sleep 1
 echo -n "."
 NET=$( ( ifconfig | grep 192.168.37 ) 2>/dev/null )

echo ""

# check if the music directory is available yet. If not mount
# the network share
if  [ ! -d /network/shared/HiFiMusic ]
 echo "Mounting /network/shared"
 mount /network/shared

# Wait for the music directory to appear
echo -n "Waiting for music directory to be available ."
while [ ! -d /network/shared/HiFiMusic ];
 sleep 1
 echo -n  "."
echo ""

# start MPD
/etc/init.d/mpd start

The Graphical User Interface

For the shiny user interface I installed GMPC – the Gnome Music Player Controller. For the purpose of starting a single program, maximised, the windows manager distributed wit Ubuntu Netbook Remix (metacity) is not ideal. So I used Matchbox, a windows manager designed for embedded devices. Start by installing it:

sudo apt-get install matchbox-window-manager

Next we create an .xinitrc script for our music user. Create /home/music/.xinitrc and add:

#!/usr/bin/env bash
/usr/bin/matchbox-window-manager -use_titlebar no &

Now we should be in the position where running ‘startx’ as the music user will not launch the whole Ubuntu desktop, but just GMPC, maximised and with no title bar.

The loose ends

All that is left to do now is to get the GUI to start automatically, and to start remuco.  Remuco is another fussy program—it won’t start if MPD is not already running. These final points are covered by editing music’s .bashrc file, /home/music/.bashrc. We add to the end of this file a while loop to ensure that MPD is running, a call to remuco-mpd and finally, run startx:

echo -n "Waiting for MPD to start *"
while [ ! -f /var/run/mpd/pid 2>/dev/null ];
 sleep 1
 echo -n "*"

echo ""
sleep 2
/usr/bin/remuco-mpd &

OK, that should be it. Restart the EeePC and it should load everything and start the GUI music player.

Remote control from another PC

The beauty of MPD is that it can be controlled from other PCs on the network. There is a wide range of client software available. I’ve found GMPC, combined with some of the big selection of plugins, suits my needs best. Ario is another good client, as is Minion, an add on for Firefox. The MPD website has a whole list of clients.

Remote control from a phone

Controlling with a Nokia E63 and RemucoThis bit’s great! Using Remuco the music player can even be controlled from a mobile phone. Remuco is a little different to other MPD clients. It doesn’t use MPD’s interface directly, but relies on a helper application, remuco-mpd, being installed on the server (the EeePC in this case). But it works well, and is really useful as my phone is always on and to hand, whereas, believe it or not, I do occasionally turn my laptop off.

The music

There’s not much point in having a quality DAC and amp and feeding poor quality music files to it. Much of my CD collection I ripped out a few years for the purpose of playing it on an in-car CD player that played MP3s. It was done in a hurry just before a journey to the Hebrides. Tinny speakers and a noisy old diesel car meant it didn’t really matter if the quality was less than perfect, so I used bitrates of between 128 and 160kbps.

On my new set up this was really noticeable, so I set about re-ripping the whole lot. For the majority I used FLAC, a lossless compression format. Towards the end I got a bit nervous about how much disk space this was using, so some of the stuff I listen to less frequently was ripped to variable bit rate MP3 using lame’s ‘extreme’ preset setting, which produces files with an average bit rate in the 200-250kpbs range. All of this was done with Sound Juicer. For the ‘extreme’ setting I had to create a new preset, using this GStreamer pipeline:

audio/x-raw-int,rate=44100,channels=2 ! lame name=enc preset=extreme ! xingmux ! id3v2mux

So how does it sound?

Great! I’ve done some entirely subjective listening tests comparing FLAC files played over this system with the original CDs played on a Cambridge Audio Azur 340C CD player. If anything, the FLAC sounds better—a touch clearer, brighter and more defined. This isn’t surprising—the 340C is quite old now and the DAC in the DAC Magic is superior to that in the 340C. Since FLAC is lossless, the input hitting the DAC Magic should be a bit perfect copy of the CD.

I’m certainly very pleased with it, and it sure beats having to change the CD every 45minutes!

Postcodes freed

On 1 April the Ordnance Survey finally opened access to data relating to the location of Britain’s 1.7 million postcodes. It’s long been a gripe of many that these data were not publicly available, and various projects have attempted to collate openly available postcode data, such as Free The Postcode and

A while ago I wrote a PHP class for manipulating postcodes, and in particular for calculating the distance between two postcodes. I wrote it for use in a utility to help people find their nearest reputable tree surgeon by searching on their postcode, which is how it came to be named FindMyNearest. Until now I’ve relied on data gleaned from open sources, often incomplete and of unknown accuracy. I used data only down to postcode district level (eg SW1A, CV32) and stored that in a delimited text file. With the release of Code Point Open I now have the opportunity to use full postcodes of reliable provenance.

The problem was that there are about 1.7 million postcodes in Great Britain. It was fairly clear that code which used a text file for storage and slurped the entire contents into memory wasn’t going to cope well with this. So I went back to the drawing board.

I’ve rewritten FindMyNearest to provide a kind of data abstraction layer. It now supports ‘drivers’ for different types of data store. The text file option is still there, but I added the ability to use a MySQL server to store the data.

Then I came across UK Postcodes, a system based on Code Point Open put together by Pezholio. UK Postcodes provides a simple API for fetching a range of data about a postcode in XML, JSON, CSV or RDF formats. I figured it wouldn’t be too hard to add a driver that utilised this, so I did just that.

And I finally got round to working out how to use Git so I’ve just put FindMyNearest on Git-hub