NAS Upgrading Fun

I’m documenting this, just in case there’s someone out there for whom it’s useful.

I had two NAS drives, both made by Synology. I have a DS210j that I bought back in 2010. It had a pair of 2TB drives in it with a mirror RAID array, and perhaps the most important thing stored on it is my offline music library. The drive was about 75% full. The RAID setup means that if one drive fails, the data should be safe on the other. I can then replace the failed drive and get back to parity.

The other drive is a DS214se bought in 2014. This had a pair of 3TB drives also in a mirrored RAID array. This was over 90% full. It’s home to a lot of my archived video projects as well as much of my substantial Lightroom photo library (essentially, everything I’m not currently working on).

Diskstation Manager (DSM) is Synology’s software that runs on these NAS drives, and it was not happy that I have less than 10%. In any event, it was clear that I needed to expand my storage in the near future. So I’d been keeping my eye out for cheap large format hard drives. I reckoned that 8TB is probably the sweet-spot at the moment in terms of value per TB. But there’s a very odd thing in the hard disk market.

If you go to somewhere like Amazon, and look for a WD Red 8TB drive or Seagate Ironwolf 8TB (both versions of their drives designed to run in NAS drives), the drives will cost you around £218-240. Now those prices can fluctuate, but north of £200 is common.

However, having spent some time in the Reddit Datahoarder group, there are plenty of folk on there who will tell you that the drives that come in some WD MyBook or WD Elements external drives are actually pretty good for using in NAS drives. Indeed at one time, they were actually Red drives. These days, they’re not strictly the same drives as you can buy on their own, but they’ll still pretty good.

More importantly, the price of these things can drop very low indeed. If you use a service like CamelCamelCamel to keep an eye out, you can stay ahead of the game. Last week, the price dropped to as low as £129 for an 8TB WD MyBook. I didn’t get onto it in time, and the price jumped back up to over £150. But then it came back down to £139, where at time of writing it still is. So I jumped in and bought two.

When they arrive, they’re of course designed for use as external drives. They come with smart plastic cases and power adaptors. Fortunately, there are YouTube videos that show you how to open these cases non destructively, a process called “shucking.”

This video worked well for me – the design in the video still being the current WD design. You just need a spudger or knife, and something to keep the tabs open. In the video, he uses guitar picks. I used a few business cards.

Then it’s just a question of sliding the case apart and unscrewing the drive from the plastic container.

Obviously you could use brute force, but that runs the risk of damaging the drive. Furthermore, if you keep the drive cases intact, you can re-purpose your old NAS drives as external hard drives. I put my two 3TB drives back into the cases as 3TB external hard drives are useful to have around. However, at time of writing, I’ve yet to work out how to format them away from Synology’s Hybrid RAID format.

Now you have your two new 8TB hard drives out. So it was a question of replacing the drives in my newer DS214se. Now I did wonder whether it was worth upgrading my NAS at the same time. As I mentioned above, my NAS is now getting on for 5 years’ old. There’s only 256mb of RAM in it, and it’s not powerful enough to do anything fancy like host 4K video and then transcode it. But in fact, that’s not a problem. I do keep some videos on my NAS for Plex (the in-home streaming platform), but I use an Nvidia Shield TV to run Plex server (i.e. run the main software) rather than run it on the NAS drive itself – something that many more powerful NAS drives can do. Instead, I just point the Nvidia Shield TV towards the video library stored on one of my NAS drives.

Similarly, I’m not doing anything like hosting a website on NAS (again – something that’s completely do-able). But I asked the question. The equivalent Synology NAS today to my DS214se is probably the DS218j. This is another £150 and in actual fact, isn’t that much more powerful than my older one. I’ll probably upgrade at some point, but not today.

Instead, I needed to replace the 2 x 3TB drives in the existing DS214se with my 2 x 8TB drives. This is relatively straightforward – but it just takes a while.

The process is basically:

  • switch off your NAS (actually – it quietly shuts itself down properly)
  • open it up (give it a good clean while it’s there – I had lots of dust in mine)
  • take out one of the 3TB drives, and replace it with an 8TB one
  • close it all up and switch on

Once it boots up, it start beeping at you. It’s noticed that one of the drives has “degraded.” It hasn’t really of course – you’ve just taken one away. Nonetheless, your next step is to “repair” it.

In essence, “repairing” it is just a matter of backing up all the data from the remaining 3TB onto the new 8TB drive.

This takes some time. 8+ hours in my case.

So I went to bed.

The next morning, there once the final couple of percent in the process had completed (you can monitor it all within DSM), it was time to switch out the other drive.

Another 8+ hours later, I was out. But the device sends an email when anything major happens. So I got the email and knew that all was fine.

At this point the system also notices that it has 8TB on each drive, so it ups the overall size available, and I was only using 35% or so of my space.

But I wasn’t quite finished. My older DS210j has been annoying me of late. While it’s not full, the discs rattle constantly – it’s noisy. Remember, you need these drives on at all times if you want to access the data, and in general, I tend to leave them on. Indeed, I can remotely access files on them if I need to.

For me, the next step was going to be copying all the data across from one NAS to the other.

The key thing with this process is not to do it via your PC (or Mac). Certainly, I have both NAS drives attached to my PC and I could do the whole thing in File Explorer in Windows 10. But that’s not smart. Instead, there’s a built in app called FileStation.

You just need to make sure that from your destination NAS (the DS214se in my case), you’ve mounted a remote drive to a folder on it. Once you’ve done that, you can then copy (or move) the files from one drive to another.

My older device actually has two Volumes on it – effectively two separate drives, although the second volume is tiny. I chose to do this one step at a time. Copying the first volume across resulted in FileStation telling me it would take 17 hours or so. A little more than making a cup of tea. And this is only an unfilled 2TB drive. But it’s a one time thing. While the estimate was initially 17 hours, the process actually took more than 24 hours. I had to leave it copying files for two consecutive nights.

Fortunately, I didn’t need the second volume – system files for a Plex server I was no longer using. So finally, I was able to retire the older NAS drive. I’m left with one NAS, and I still have about 46% of space left.

A final note is to say that I sync my Lightroom library to Amazon Photos – mainly because it’s an inexpensive bolt-on for Amazon Prime members, giving unlimited storage for photo files (including RAW photo formats). What it doesn’t include free is video.

I used to use Amazon Drive to back up both NAS drives completely, when Amazon Drive allowed unlimited storage. But that’s gone. I have to pay a modest extra fee for all the sidecar data that Lightroom stores. Non destructive editing requires a lot of these files. But it’s not too bad.

My ideal would be to securely backup all my files offsite. This is very do-able, and there’s an app within DSM precisely for that. However the costs begin to mount up. For 4.5-5TB of storage, you’re looking at £/€300-350 a year. Getting the data back might increase those costs. Amazon, for example, has Amazon Glacier. I assume that it’s stored on a tape format. The idea with Glacier is that you don’t need to get the data particularly regularly – just know that it’s there. That’s as well, because recovering the data is actually quite expensive.

In my instance, this is really about securing photos, music and video – with a few documents thrown in. If my flat where to burn down, or someone steal my NAS drive, then I’d know I could recover everything. But doing that in a cost-effective manner is tricky.

One solution is finding a friend with a big internet pipe who’s willing to let you put a second NAS drive in their place. You could then keep the two in sync. You could obviously “host” a NAS drive for your friend in your house at the same time.

A previous “off-site backup” solution I’ve used in the past is to simply buy another external hard drive (perhaps one of those 8TB WD MyBooks!), back the NAS up to that, and then keep that drive securely in your workplace. If you then brought it home once a week/fortnight/month, you could update the backup and return it to work the next day. You’d never be more than a week/fortnight/month behind if the worst happened. Indeed, you could buy two drives and shuttle them back and forth – keeping one at home and one at work at all times. But that may be over-kill.

I realise that this isn’t the most exciting thing to write about. And when I talk to people, I realise that relatively few people really look after their data. If you don’t have all that much, you can perhaps rely on cloud storage services. Google Drive, Dropbox and iCloud all give you a bit of free space, and will charge you for more. OneDrive can be good value if you want Microsoft Office. But many of these max out at 1TB in space. Even Google, who used to be quite generous with Google Drive space, giving it away with Android phones or Chromebooks, seems to be more inclined to charge these days.

Your parents’ or grandparents’ photo albums may seem quaint, but those albums will be much safer than a lot of photos that are only stored safely while users hang on to their social media accounts…

Longer Lasting

Over the last few weeks, a lot of things around me have started to break.

Most annoyingly, the TV that was given to my father just over 18 months ago has developed a fault in the panel – the lower quarter flickers and then stays slightly greyer and fuzzier than the rest. It’s only 18 months old, but it’s outside the warranty period.

At the same time as the TV, they got a soundbar as well, and that has now broken too. I simply don’t understand how, and I’m really not inclined to buy another Sony A/V product right at the moment.

Meanwhile at home, my Sony receiver is also playing up. It will turn itself off at random intervals – normally after a few hours. The TV no longer controls the sound on it. I’m not 100% certain when I bought it, but it’s well over 5 years old at this point.

Then there’s my Sky box. This has on occasion needed a reboot every month or so, but now it’s getting a little flakier. It seems to lose signal from one and then the other LNB at various points. A cold reboot fixes it, but the end is clearly nigh.

Most urgently, there’s my washing machine. A week or so ago I had completed a wash cycle, but had not yet emptied the load. While I was doing some washing up, I heard a loud bang, the source of which I couldn’t track down. I subsequently emptied the load and on trying to put a second load on, noticed that there was no power. The fuse in the machine’s plug had blown. But when I replaced it, although it seemed to come back to life, a flashing door icon restricted me from doing anything else. Nothing I did could make the machine work.

A friend on Twitter suggested that replacing the master board on a washing machine isn’t that difficult. I quick check of a YouTube video confirmed that such a job was well within my wheelhouse. However, I’d be spending £40-50 on a part that I didn’t know wouldn’t instantly break again.

Suffice to say that a new machine arrived on Friday. And to give it some due, the previous machine had run without issue for 10 years.

The reality is that most of these devices aren’t really fixable. A good washing machine technician could probably have sorted out that. But once call-out, parts and some billable time had been included, I’d be at at least half the price of a new machine. Things with moving parts do break after all. And this new machine will be more energy efficient than my old one.

The TV and soundbar faults are more frustrating. I can’t attempt to fix them, and again it’ll be cheaper to buy new items. As an aside, I’m giving Resolver a chance to help to me. But it gets complex because I don’t live near my parents.

On a related note, I started looking at Sky HD options considering it looks like the box is going to die soon. Sky has its new Q box, but goodness – they’re expensive.

One significant issue for me is that my current Sky box is quite full. I would need to “watch down” everything before switching devices, as there’s no supported way to transfer recordings between boxes (There are very much unsupported options of either using HDMI out to record to another device, or using software to decode programmes direct from the machine’s hard drive).

More to the point Sky Q boxes are rented from Sky and not owned. There’s a flat £12 a month fee for that rental, which quickly adds up. Plus they seem to want £199 for installation – as I’m neither interested in, nor wish to pay for multi-room. Of course, if I was a new subscriber, I wouldn’t have to pay nearly as much as this.

Since I don’t have a 4K TV, this is all a bit moot, since it wouldn’t offer a great deal more. Instead, I can buy a new Sky HD 2TB box for around £140 from Amazon.

Finally, over Christmas, I noticed that my laptop’s WiFi was a little flaky. I mostly use it at home, and connect it via ethernet so hadn’t noticed. Fortunately, that at least looks like it’s an achievable fix that I can do myself.

YouTube: Different Things to Different People

There have been a few stories kicking around about YouTube’s end of year “Rewind” video. YouTube has been making these for a few years, and they’ve grown to become big-budget affairs.

But this year’s video is seemingly the most “disliked” video in YouTube history.

Now I should say upfront that I have not watched this video. Because frankly, the part of YouTube it represents is not how I watch YouTube. Of course I’m well away from the 15-24 demo that this video perhaps is targeted towards, but that doesn’t mean I don’t watch YouTube. I would say that it’s probably my go-to video streaming app. I might watch it in my browser; perhaps on my phone; but definitely on my TV via my Nvidia Shield TV.

I’m not saying that teenagers don’t watch larger quantities than me, but I suspect that the Venn diagram of my YouTube viewing and those who’ve taken the time to dislike the YouTube Rewind video looks something like this.

Using proper notation, A ∩ B = Ø.

But what this doesn’t really show is the size of the universe of YouTube videos. YouTube has 1.9 B users a month, and they’re not all watching the same videos.

Put that another way. Last week ITV aired the series finale of I’m A Celebrity… and over 11m people watched it. But 56m watched any TV last week. 

In other words, 4 out of 5 TV viewers didn’t watch I’m A Celebrity… That’s not to belittle a sizeable TV audience, but to point out that just because you and all your friends have been watching, it doesn’t mean that the entire population is watching.

Scale that up to the rest of the world and YouTube, and you begin to realise that YouTube for a Rewind-featured YouTuber is very different from YouTube for, well, me.

I’d love to know what the combined 2018 views of all the YouTubers featured in the Rewind video is as a percentage of all the video views on YouTube in total. I’d be willing to bet that it’s relatively small.

As for my viewing habits? I’m watching tech review videos, Global Cycling Network videos, innumerable tutorial and how-to videos and so on. I’m not watching vlogs or anything with a clickbait-y title. I’m also not watching a great deal of music, whereas others are watching an enormous amount of music. 

Different folks, different strokes. 

I bet your viewing habits are quite different from mine.

So when I read these stories about how disliked a video is, I think that there are lot of people who are in a big bubble of their own design and that it’s largely because devoted followers of a big YouTuber who posts the odd link to nasty right-wing sites, was not invited to participate in something YouTube uses to celebrate the platform and so his fans took revenge.

IKEA Tradfri Lights Working with Hue, Alexa and Google Assistant

Here’s another one of those tutorials that I really shouldn’t need to write but somehow do, because despite dozens of tutorials already existing online, it was only a combination of things that got everything working for me. In short, I already have a Philips Hue lighting set-up and I wanted to add some IKEA Tradfri bulbs to it.

I was in my local IKEA when I noticed that they had Tradfri E14 bulbs. Tradfri is IKEA’s smart lightbulb range, introduced a couple of years ago, and it has become a cheaper option for automating your lights.

The main source of light in the rear of my living room takes 5 E14 bulbs, and although the front half is already nicely Hue controlled, it was always just too expensive to do the rear. A twin pack of white bulbs costs £40 on Amazon at time of writing, or £25 for a single bulb. So I’d be looking at over £100 to make one lighting fixture Hue-controlled! (OK – other options include replacing the light switch, fitting a new light to my ceiling that takes fewer, cheaper bulbs, or just continuing to use the light switch like a normal person).

The key thing here was that the IKEA bulbs are £7 each. £35 (5 x £7) is still a lot, but it’s much more palatable than £105. Sidenote: I suspect that there is some kind of economic theory that explains why I’m happy to pay £35 to automate a light, having previously worked out £105 as being the ‘regular’ price.

Now, over the last year or so, IKEA’s Tradfri range has become more compatible with Philips Hue. Both use Zigbee to connect together, but IKEA’s bulbs are designed for it’s own set-up out of the box, and I didn’t want to invest in that. In any case, I think with one or two hiccups along the way, Hue’s app and connectivity is excellent.

So in theory, it should have been simple. Read a couple of guides online. Watch a YouTube video and away you go.

Well it wasn’t quite that simple.

One of the key things I read was that when connecting a Tradfri bulb to an exiting Hue set-up, you should first power down all your existing Hue lights. For precaution, I also unplugged a couple of Meross power adaptors that I use (these tend to appear relatively inexpensively on Amazon, and are useful for controlling things like desk lights and fans).

Then you need to do a reset of the IKEA lights by flicking them off and then on six times. This should result in the bulb flickering slightly to show you that it’s ready. Despite multiple attempts, I couldn’t discern any flickering. I felt that I was more likely to blow a fuse than get the bulbs into reset mode.

Nonetheless, I pressed on.

Then you need to make sure that the lit bulb and the Hue hub are within 30 cm proximity of each other. On YouTube people tend to use a lamp to temporarily set up their new bulbs. They dutifully place their lamp with Tradfri bulb close to where their Hue hub sits. But I don’t own a lamp with the right E14 screw fitting. So I had to fit a long network cable to my hub which fortunately is not too far from the router or the light in question. The hub’s power cable just about stretched too. This is all worth knowing in case you’re using an exotic bulb type and can’t place the hub near the light fitting in question.

Unfortunately, I repeatedly failed to get the bulb to connect using the Android app. Simply searching for a new bulb should have found it. This was frustrating.

The Solution

I ended up using using an app called Hue Lights. It’s iOS, PC and Mac only, and is not made by Philips. So I used the iOS version on my iPad. And I carried out the process one bulb at a time.

Then flicking the light switch off and on 6 times, holding my Hue bridge to the bulb and using the Touchlink function in Settings on the Hue Lights app, I pressed “Force bulb to join bridge.”

Almost immediately the bulb visibly pulsed, and after an arm-aching minute or so (Remember, I was holding my hub to the ceiling), it was properly set-up. I repeated this 4 more times, one bulb at a time, removing bulbs I’d just added to avoid any bulb “confusion.”

Then I replaced all the bulbs, switched everything back on, and the bulbs now showed up in the Hue app on my Android phone. I could allocate them to a new “Room” and control them properly that way with the Hue app.

But for some reason, none of them were yet being recognised by either Amazon Alexa or Google Assistant. At first I thought it might take a few minutes for them to register, but time passed by and none of it was controllable via a smart speaker.

In the end, it required me to disable and then re-enable access to my Hue account from within both the Amazon Alexa app and the Google Home app. 

Once that had been done, all the new bulbs showed up, and I was able to set them up into various groupings and rooms as makes sense within each of the apps. It’s worth spending some time thinking about which combinations of lights you want turned on simultaneously. I think that the Alexa app is better for this than the Google app, but you can fight through and eventually get things sorted in Google’s app too. 

What all of this has told me is that adding off-brand bulbs is not for the fainthearted. But then, I don’t think setting up smart homes is especially easy in any case. And as I noted with my recent piece about setting up radio alarm clocks, I don’t think that the apps are as user friendly as they could be just yet.

Still, if you want to save some money, then IKEA’s bulbs are a decent enough way to go, and aside from automating my living room lights completely, I’m now able to dim some lights that previously weren’t dimmable which is a nice bonus.

Building a Desktop National Rail Dashboard

Building a Desktop National Rail Dashboard

For ages now I’ve been meaning to put together a home dashboard: a screen that gives me up to date information about my local railway station with the departure boards showing.

A slightly bigger ambition was to have the dashboard also display times for nearby buses – and perhaps the weather and Twitter news feeds. 

It should have been a relatively simple project, but it seems that live dashboards are either bespoke paid-for things, or are no longer properly developed.

Either way, I decided to do something relatively simple. There was something called which is no longer maintained. In its place is a fork of it called smashing. But I couldn’t get on enormously with either.

What I specifically wanted was a dashboard that would be Raspberry Pi powered, and could make reasonable use of a small screen.

Some time ago I bought the parts, but it has been one of those projects that I started and failed to complete. So to get things going, I built a simplified version using the parts I already had:

  • Raspberry Pi – I used a 3 Model B+ although the newer slightly smaller and cheaper Pi 3 Model A+ would probably work.
  • A screen – I used an official Raspberry Pi 7″ Touchscreen. It’s only 800×600 and it’s a shame it doesn’t come at a slightly higher resolution.
  • A micro-SD card.
  • Optionally, I also used a Pimoroni Touchscreen frame to mount it all together. They come in a range of colours.
  • A Raspberry Pi power supply. You’re probably best using a full 5.1V supply since there’s a fair bit of pull to power both the Pi and the screen together. Don’t rely on a cheap old phone charger. That said, mine is powered by a USB port of an Anker 60 W charger.
  • You’ll also need a USB keyboard to set things up initially. The touchscreen will suffice for a monitor.

Now in another world, I’d design something a little smarter and use some nifty CSS to make everything look nice. But as you can see from the above picture, I didn’t do that.

The hardware is relatively straightforward, and in this instance there isn’t really any software to create or install. But there are still challenges!

I’d found a handy link on a Reddit forum that presents a station departure board, hosted by National Rail.

This is the page for Highbury & Islington station in London.

I strongly suspect that this link is using someone else’s API key. I suppose I should really get my own. But it seems to be working fine. As well as a User ID, there’s space for the all-important CRS code and a mysterious H value which I’ve not quite worked out.

The key thing with this is that you need to find the CRS code for the station you’re interested in – the three letter code every station in the UK has. The list of codes can be found here.

While the page is a little underwhelming in design, it does scroll if needed and has neat little features like telling you where the train was last reported, so you can live track your train.

So I had a basic website that I wanted to load up. Now it’s simply a question of launching the Pi on boot, straight into Chromium (the Pi’s default browser) and show the page fullscreen – or in so-called ‘kiosk’ mode. There won’t be a keyboard attached, and the idea is that the device does one thing, and does it well.

This seemingly trivial task took me an annoyingly long amount of time to work out, because lots of people do it lots of different ways, and it seems that some methods no longer work with the current default installations of Raspbian, the main OS for Raspberry Pis. If you’re a Linux expert, then much of this might be trivial. But I’m not, and for me it wasn’t!

That all said, in the end, by carefully following the steps on this page from Die Antwort, I got it all to work using their methodology. I can confirm that this works on the current version of Raspbian (the main operating system used on the Pi) – at least at time of writing.

Note that if you buy the Pimoroni frame as I had, then you’ll want to follow these instructions which tell you how to rotate the screen 180 degrees by default, since everything will otherwise be inverted. You can do this early on after installing the Raspbian operating system below; the precise orientation you choose will depend on how you want to place the screen.

The key thing here is to only install Raspbian Lite, which doesn’t actually include a graphical user interface. But if you go step by step through the Die Antwort instructions, you should get there.

A few things to note for those who are a little uncertain:

  • Once you’ve downloaded the Raspbian Lite operating system, you’ll need something like Etcher to flash the OS onto your micro-SD card. Pop the micro-SD into your desktop computer and point Etcher at the downloaded file. Don’t worry if it’s a *.zip file – Etcher will handle that without you needing to unzip the file yourself.
  • It’s worth turning on SSH when you set-up Raspbian for the first time, so that you can do most of your work remotely from another PC. If you’re not too familiar with this, it means that a program like PuTTY (on Windows) will let you remote into the Pi in console/terminal mode. That makes it easier to copy text from website guides straight into configuration files and the like. You will need to work out your IP address to do this. Type sudo ifconfig on the Pi to find out your IP address. It will probably next wlan0 and be of the form 192.168.1.x. These represent internal-only IP addresses.
  • When you’re setting up WiFi, you need to know and spell your WiFi SSID correctly (That is, your router’s name). Unlike in graphical interfaces, you just type in your WiFi hotspot name rather than select it from a list of available ones. Type carefully! You also need to set your region so that the right WiFi frequencies are searched.
  • When it comes to editing files, the Die Antwort instructions often don’t explicitly say that to edit a file, you should preface the file with sudo nano. e.g. sudo nano /etc/xdg/openbox/autostart in the ‘Openbox Configuration’ section. “Sudo” means act as a superuser or administrator, and “nano” is a basic text editing program in Linux.
  • When you try out launching the screen, do so on the device itself and not via SSH. You can do everything else remotely, but you’ll want to plug a keyboard into the Pi to test when the instructions invite you to type startx — -nocursor.
  • The thing that threw me most was in the section ‘Start X automatically on boot’ – in particular, I didn’t know where the .bash_profile file was to be found or stored. What you need to know is that it doesn’t exist initially and you create it by typing ‘sudo nano .bash_profile‘ from the default location.
  • Nano is a real old-school text editor. But if you’re copying and pasting via SSH, it’s worth noting that right-hand click is paste rather than Ctrl-V as usual. To save it’s Ctrl-X, then Y for yes, and then Enter to save and close. If you’re typing directly into the Pi then type slowly and carefully.
  • Don’t forget that you need to replace http://your-url-here with the link to your website. In my case this was a long link beginning…

One further amendment I needed to make to the Die Antwort instructions was to shrink the default Chromium zoom size a little. Because the Pi’s touchscreen is only 800×600, everything looked a little squashed on the screen. After experimentation, I chose 70% as an appropriate size to squeeze a bit more text onto the small screen. To fix this you should add some extra text to the Openbox configuration to scale the browser down, using the delineator –force-device-scale-factor=[decimal]. ie. for 70% of full size use:

chromium-browser --disable-infobars --kiosk --force-device-scale-factor=0.70 'http://your-url-here'

When you test it’s also working, remember that you can break out of the page by pressing Ctrl-Alt-Backspace with a keyboard plugged into the Pi.

If you follow all these instructions carefully, you should end up with something similar to me.

This all gets me up and running, but there are lots more improvements to make!

  • One additional one is the idea of cycling through several screens with different information. E.g. I might add a separate tab to Chromium with bus times. Then I can automatically cycle between the tabs to show different screens. There seems to be a route to doing this in these instructions. Note that other aspects of this page won’t work with the solution employed here.
  • The biggest thing I’d like to change is not have the screen on the whole time. Indeed, during set-up we go out of our way to turn off the screensaver. Ideally I’d love to turn off the screen at certain times and turn it back on at others. This screen is most valuable for me during weekday mornings and at weekends. Outside those periods, I would be happy for it to be turned off, or just be awoken by touching the screen.
  • Changing the look somewhat would be nice. Some personalisation of the style would be good.
  • I think the South Western Railway logo in the photo above is related to the username I’m using. 
  • It would be great if I could load fewer departures on this size screen. Note that because I’m using a touchscreen, it is scrollable. But I’m not looking at a major London terminus so I only need a handful of departures at any given time.

As I said at the start I had previously had bigger ideas about putting together a much more comprehensive dashboard, but aside from the touchscreen’s low resolution not being really suitable for that, you’ll notice in the background that I also have a Google Hub, and to be honest, that takes account of a lot of my needs for things like time and weather (as well as photos!). That said, I’ve still not worked out a simple way to quiz Google Assistant about train departure times. On the other hand Amazon Alexa is great when you set up the skill for your “commute”, but is painful to use for anything else. So creating this departure board is actually a useful exercise.

Hopefully this will be of use to someone else. I’m pleased with the result. It could be better, but it does the job.

Surface Go – Review


In 2015 Microsoft released the Surface 3 and I bought one. It wasn’t the most powerful Windows PC ever. Indeed, it was very much under-powered. But it was light, portable and ran full Windows applications. I bought it to allow me to run full Windows applications when I was travelling. Although it had limited on-board disc space, I managed to use both Lightroom and Adobe Audition on it without problem. I would never have wanted to run anything too powerful, but for doing some edits on-the-go, before importing my temporary Lightroom catalogue into my main one when I returned home, it was fine.

While tablets are great, and Chromebooks are wonderful, they simply can’t do everything, and in particular run more powerful applications like Lightroom and Audition. Even with the inclusion of Android apps, in my experience, they’re not enough. 

So the arrival of the Surface Go was intriguing for me. I’ll say at the start that I certainly don’t think this device is right for everyone. If you’re more of a consumer rather than a creator, then I suspect an iPad is your best bet – perhaps with a keyboard attachment for emails.

The previous Surface 3 had an Intel Atom processor which meant it was slow. However, even then, I could run Lightroom and I certainly edited audio quite happily in Audition. I was even able to run bespoke software that crunched radio ratings figures in a fairly acceptable timeframe. It ran Office applications at an acceptable speed. 

Given that the Surface 3 basically got me over the line with my needs, why did I feel the need to update? Well it was by no means perfect. The keyboard was below average and the battery power was very iffy. On a recent trip I would find it randomly dropping from 60% to 0% in a few minutes – essentially shutting down on me. I’d be scared to take it anywhere without a wall-socket. It was also very slow. Slow to boot and slow to launch applications.


Microsoft sells the new Surface Go in two versions: the headline £379 version, and the version that you should actually buy at £509. That extra cost gets you double the RAM – 8GB instead of 4GB – and double the drive space – 128GB SSD instead of 64GB eMMC. I can’t think of a reason why anyone apart from the lightest of web-user would want the cheaper model.

The processor is an Intel Pentium Gold 4415Y. I confess that the name ‘Pentium’ concerned me a little, since it’s not a processor that I’ve come across since my old desktop days. But it’s a Kaby Lake era processor with dual core. 

The actual device itself feels very premium indeed. A lovely solid aluminium body with a kickstand that folds out to your preferred angle (the Surface 3 had three fixed positions, limiting your choices). The screen is gorgeous and of course touch sensitive. It’s an 1800×1200 panel covered in Gorilla Glass 3. So not quite full HD, but then this is a 10 inch device, so that would really be overkill. 

I bought one of Microsoft’s £125 Type Covers. There’s a £100 black model and then a few choices of colours which use Alcantara – a material that does feel nice on the hand. As with previous Surface devices, a powerful magnet ensures that the keyboard and Surface Go are locked into position properly. To separate them, you just firmly pull the two apart. They keyboard is a significant improvement on the Surface 3 keyboard in that the touchpad is bigger and better, while the keys are firmer, as is the case itself. Given the size of the device, this is a small keyboard. But it doesn’t take long to get up to speed with it. 

The only slight negative I have is that it’s not a backlit keyboard – my Surface Go keyboard was. [Update: Ignore this. The keyboard absolutely is backlit.]


I’ve yet to buy a new Surface Pen because the one I used with my previous Surface 3 works absolutely fine. In truth, I didn’t use that pen a great deal, because it wasn’t a terrific device for taking notes or trying to draw with (I am not an artist). However, the slightly smaller form factor of the Surface Go means that writing on it is more achievable. But I confess that I’m still looking for that perfect note-taking app. I’d also like a “learn to draw” app while I’m at it!

You can launch one of a selection of apps by clicking the button on the top of your pen. The app that allows you to annotate a webpage can be especially useful. But there is minimal lag when writing on the screen, and Microsoft’s in-built handwriting-to-text converter is pretty decent. How well it works for you will partly depend on your handwriting.

The front facing camera on the Surface Go has built in Windows Hello, meaning that you can use it to unlock the device. I found it to work superbly in even very poor light. 

The Surface Go comes in Windows S mode – a limited version of Windows that only allows you to install applications from the Windows Store. This is of no use to me as I wanted to install Chrome and applications from Adobe’s Creative Cloud, neither of which are in the Store. Fortunately it is relatively simple to switch to Windows 10 Home, and I was happily installing the apps I wanted. 

I found that Chrome runs seamlessly and I detected no slowdown or delays even with plenty of tabs open. Microsoft Edge is a very ‘needy’ application of course, and you’ll get plenty of notices suggesting you give it a go, but Chrome works very well across multiple devices for me. 

But perhaps more usefully, here are my experiences of running more powerful Adobe applications. It’s the ability to run these apps that made want to still have a portable Windows machine that’s powerful enough to do some creative work.

Adobe Audition – Unless you’re in radio or podcasts, this might be a bit niche, but it’s an application that I use regularly on the go. The Surface Go accomplishes this with ease – although I should note I probably don’t use more than six tracks simultaneously. Your mileage may vary if you’re creating a 128 track masterpiece with lots of processor intensive 

Lightroom Classic CC – This is perfectly useable. I was able to use Lightroom on my old Surface 3, so this didn’t come as a surprise. Again, it depends what you’re trying to do with the application. As mentioned previously, I create small temporary catalogues on the Surface Go, and then import them into my main catalogue when I get home. The application launches fairly fast, and it really only slows down if you’re trying to do a lot of batch processing – e.g. applying a filter to every photo you import. Your bigger concern will be how many RAW files, for example, you can store on the machine while you’re using it. It’s entirely possible the cards in your cameras will be larger than the space available on the Surface Go.

Photoshop CC – You can actually use Photoshop reasonably well. I wouldn’t recommend it on a mega-pixel image. And obviously there are some screen constraints (although these go away if you plug in an external monitor). I wouldn’t want to do anything too intensive – 3D for example. But some quick editing of images on the go, it’ll work just fine. The slowest part of using it is actually launching it. In my tests, it takes around 30 second to start up. It’s worth noting that Adobe has a Touch interface for Photoshop, with touch gestures as well as support for the Surface Pen.

Illustrator CC – This is an application that you might well want to use if  you’re artistically inclined. As with Photoshop, there’s a Touch interface which by default launches when you remove the keyboard. Again, I found Illustrator useable on the Surface Go. I suspect that other lighter applications, such as the pre-installed Sketchable are better suited for pure drawing (the app is free, but many premium features need to be unlocked at a cost of £24.74).

Premiere Pro CC – This is probably at the extreme end of what I’d want to attempt with a machine of this power. It’s clearly not designed for high-end video editing. But you may be out and about, and want to put something together. You might want to think twice about editing 4k 360 video on this, or anything. As with Photoshop and Illustrator, it takes a while to launch. As with other Adobe CC apps, Premiere takes about 40 seconds to load, and I did run into a difficulty. It seems that the Open CL GPU Acceleration on the supplied Intel video driver is incompatible with the current version of Premiere (and After Effects). This is instantly noticeable because it my tests, the video was streaked in green and pink. The workaround is to enable the Software Only playback engine. That’s probably not great long-term solution however. With any luck either an updated Intel driver and/or an updated version of Premiere (which is due to be released within the next week or so) will solve the problem long term. I did look for updated drivers on the Microsoft site, but couldn’t even find Surface Go drivers listed.

That all said, editing a 30fps 4K video from my phone was fairly painless, and for quick edits, I’d be happy to use this machine. Again, I would avoid anything too processor intensive such as stabilisation effects with Premiere, but otherwise, it’s very usable. The bigger problem you’re likely to face is having much storage space for video editing. Video will eat up the limited space very quickly – especially on the 64GB model. One option would be to use a USB-C SSD such as the Samsung T5 range. 

I didn’t try After Effects because that would have been daft to attempt. I suspect that this is the most processor hungry application I use, and it can sometimes struggle on my i7 enabled Dell XPS with 16GB of RAM. That said, I would expect the same short-term video issue as above.

NB. I’ll let others tell you whether or not you can play games on this machine. Personally, I’d be looking elsewhere for a gaming machine. I’ve done some very occasional PC gaming in the past, but this wouldn’t be my go to. 

There are two other things you’ll probably want to buy alongside the keyboard. The first is some kind of case. I ended up with a case that was designed for an iPad Pro. I bought a Tomtoc 10.5 inch sleeve which fits the Surface Go perfectly. It has a super-soft interior which should fully protect your device, and there’s a zip pocket at the front which is able to accommodate a Pen (assuming you have the older non-magnetic one like me), and a dongle. 

The great thing about the Surface Go’s size is that even in this case, it feels very small, and will happily slip into all but the very smallest of bags. Even “fully-loaded” with the Surface Go, case, pen and dongle, the weight only comes to 1020g. The Surface Go on its own, excluding the Type Cover, is 522g, but I think 1020g is a more realistic “real world” weight that you’ll be carrying around.

Dongle? Yes – this device comes with a micro-SD port (into which I put a 128 GB micro-SD card for additional storage), and a single USB-C connector. While the Surface Go comes with its own bespoke Surface charger, with a unique connector, the device can also be charged using a USB-C charger. In practice I found that my Pixel 2 charger worked absolutely fine. That’s really useful if you want to travel light. But that also means that to get USB-A ports (e.g. for USB dongles), or external video connections, you’re going to need a dongle. I bought a Lenovo dongle that had just about the right connectivity options for me at a reasonable price. In my case, that was SD, micro-SD, USB-A and HDMI. It also has a USB-C through power connector. Pricier versions are bigger but include Ethernet and even VGA connectivity, which might be useful when presenting on the road. Note that the USB-C port is just that. It’s not a Thunderbolt 3 port. So don’t think of plugging an eGPU into it (You wouldn’t want to do that anyway, but it’s worth saying).

Still, I won’t complain too much about the USB-C inclusion. Microsoft has since refreshed the rest of its high-end Surface line, and none of those devices include so much as a single USB-C port! 

One issue I did have with the Surface Go was during my initial Windows 10 set-up. My finger slipped on an early dropdown, and I selected Germany rather than the UK. This then sent the set-up into German, and didn’t allow me to go back and correct my mistake! I couldn’t escape out of it, and had to complete my set-up in German. Then, although I should have been able to change language to English, I didn’t fully get the machine working in English. Some system messages still came up in German. 

After fighting with this for a while, I ran a Windows ‘Reset’ which after some time, put my machine back into it’s initial state. This time I managed to click the right country and set-up continued unhindered. 

This is more a problem with Windows 10 rather than the Surface Go. Why no back button?

Battery life seems good. Microsoft claims 9 hours of video playback, but I’m wary of too many claims on battery life since it’ll clearly depend on what you’re doing and how old the device is. But I think that there’s probably enough battery to get you through a non-intensive day.


This thing is small and light. You can put it into just about bag, or even a large pocket, and not really notice you’ve got it.  The screen is excellent, and the keyboard is very usable – a step up from that on the older Surface 3.

The Surface Go is very capable at most basic tasks. Microsoft Office runs well, so for those for whom either the online versions of Office or Google Docs aren’t quite enough or are too reliant on connectivity, this will do very nicely. I’d say that it’s perfect for travel, fitting neatly onto fold-up tray tables on trains and planes.

You can actually run some quite power-hungry applications with relative success. And Windows Hello is amazing – even if I’m not looking face on to the camera, it recognises me, and logs me straight in.


You do need to buy a few accessories. First off, there’s the keyboard. Since the whole device is smaller than the Surface 3, my Surface 3 keyboard doesn’t fit – the connectors are different. So you’ll need to spend another £100-£125 for that. And if you want to use a mouse, be sure to get a Bluetooth model – not the more commonly available wireless mice that use a USB-A dongle. I found a nice Microsoft Surface Mouse in a matching cobalt for £30 at Argos.

I wish there was more on-board storage and that Microsoft had made 128GB and 256GB models. But you can use a micro-SD card which should be sufficient for video and music download needs.

You can obviously get cheaper laptops that are more powerful than this. But they’re not as portable. So getting a Surface Go is very dependent on your specific use case.


If you want a highly portable full-fat Windows laptop for browsing and Office-type tasks, then this is the perfect machine. It really isn’t much bigger than an iPad, but with that much more flexibility that full Windows applications allow. Throw it in your bag and go.

Obviously if you don’t need fully featured Windows applications, then an iPad and keyboard might work better for you (I remain unconvinced about the need for an iPad Pro). But for my use case, which involves a wider array of applications from Office to Adobe’s Creative Cloud, this actually works really well.

It wouldn’t be my primary PC by any means, but as a portable addition, it’s excellent.

Of note: Microsoft subsequently announced its new range of Surface laptops and an upgraded Surface Studio targeted at designers. These are respectively Ultrabooks and aimed at graphic designers and artists. As such, they don’t really compete, and are not comparable. In any event, as things stand, I prefer Dell’s XPS line to Microsoft’s Surface line for this class of PC. Asisde from anything else, they come with USB-C, as the Surface Go does!

Google Podcasts

Without an enormous amount of fanfare, Google yesterday launched Google Podcasts for Android yesterday, with the possibility of being game changing. I’ve long argued that for the Android/iOS podcasting gap to be closed, Google needed to get involved and create a generic app.

Apple Podcasts is a pre-installed app on every iPhone sold, and with strong backing of podcasts from the outset via the iTunes store, Apple users have consumed podcasts at a far greater rate than Android. Even today, with iOS share slipping slightly, the proportion of podcasts consumed by iOS devices is massively out of kilter with smartphone ownership. In most countries in the world, there is a higher Android user base than iOS.

All of this means that, unless we somehow infer that your choice of smartphone is a strong indicator for how you listen to audio, then there is a massive untapped Android market out there.

Previously Google has only played a little in the podcast arena. They added podcasts to Google Play Music. But only in the US. And podcasters themselves had to add their podcasts into Google Play Music themselves. A combination of those two things meant that that ex-US podcasters who wanted to list their podcast with Google had to go out of their way to employ VPNs to even get their podcast registered. Furthermore, Google Play Music cached audio meaning that podcasters couldn’t see a comprehensive picture of their podcasts’ performance across a range of platforms. Furthermore, newer technologies like dynamic advertising wasn’t possible. The advert baked into the podcast when it was captured by Google remained there in perpetuity.

Google just wasn’t taking podcasts seriously. But that was obviously changing and when Pacific Content published their series of articles on Google’s new podcasting drive earlier this year, things Google had been doing began to come to light. Although the scale of podcasting continues to grow, with more people and organisations releasing more podcasts, and more revenues being derived from them, it was perhaps the growing importance of audio to Google itself that has really pushed things along. Google’s Home and Home Mini devices have been massive sellers, with the company locked in a battle with Amazon’s Echo for grabbing market share in Voice (Despite Apple’s Siri being first to market, Apple is playing a massive catch-up game in this market).

Voice control has come to be an important way we interact with technology with both our phones and our devices in our smart homes. Machine learning has meant that voice comprehension and contextual analysis has rapidly improved. And from there music and speech are perhaps growing in importance. So podcasts fit in neatly.

All of this explains why Google’s new podcast app, isn’t actually an app at all. It’s really a view of Google Assistant. For quite a while now, you’ve been able to ask your Google Home device or your phone to play a podcast. This “app” therefore just makes this a little cleaner.

In fact the app is actually pretty basic. The average podcast app you can download on the Play store is likelier to be much better featured than Google Podcasts. Even something as basic as downloading podcasts for offline listening – the absolute bare minimum you need for any podcast app – requires you to change permissions in a truly bizarre way. Instead of getting a pop up permissions dialog box as you’d expect from recent Android iterations, you’re taken to a user-unfriendly App info page where you have to choose Permissions and then turn on Storage. It really isn’t very obvious, and I suspect many will fall at the first hurdle.

The rest of the app is very basic. The “Top Podcasts” are all very obvious and popular US ones: This American Life, Serial etc. And then all the usual suspects are in each of the category selections. The only two non-US podcasts I saw were the BBC’s World Cup Daily and The Guardian’s Football Weekly podcast. There was a Five Live section for me, which may have been because I subscribed to a Five Live podcast through the app in testing.

Now to be fair, this isn’t necessarily a terrible idea to highlight the podcasting big hitters. If you’re just discovering podcasts, then you probably want to listen to all the favourites. And equally, I don’t really know of any app that is very smart at selecting podcasts for you. Indeed, for all it’s revered elsewhere, I find even Netflix misses much more than it hits with selections for me.

Obviously a key benefit that Google Podcasts does have is that if you start listening on, say, your Google Home Mini and then leave your house and listen via your phone, you can carry on where you left off. But in the time I’ve tried the app, I’m unlikely to leave PocketCasts as my podcasting app of choice, which also lets me move between phone and its desktop web app. For smart speakers, I tend to use Cast to keep things in sync and stay on top of which episodes of which podcasts I’ve listened to. It has other much deeper functionality that Google’s offering lacks. This is probably purposeful on Google’s part, and other app developers will probably be relieved.

None of this is to say that Google Podcasts isn’t very important. Any podcast creators should build links to Google Podcasts as soon as possible, include their badges and generally make sure they’re listed correctly. Podnews has a decent FAQ about what you need to do. At the very least, when people share a podcast socially, they can now include a Google URL as well as an iTunes one (NB. They should still really share a link to a website where a range of options are available including the podcast’s unique RSS feed).

However, I’m not sure this is going to be quite the game changer it might have beene. I don’t see the app being pre-installed on phones, and I suspect that most of those who’ve installed already are those who are already very familiar with podcasts. Yes, it’s true that the podcast functionality will be pre-installed in that it forms part of Google Assistant. But it’s not clear that Google is pushing a page as a destination, in the way you might go to the YouTube homepage to see what new videos have been published, or you would open Spotify to purposefully listen to music.

That said, podcast usage is going up – there are some good global numbers in the most recent Reuters Digital News report (Interestingly, the UK is at the lower end of the range with 18% listening to podcasts a month. In South Korea for instance, it’s 58%!), and this initiative can’t but help drive that listening upwards.

One really interesting area Google is planning to tackle is the idea of creating subtitles (or captions) for podcasts using Google’s AI. Relatively few podcasts have transcripts of their programmes, and that makes searching the content within them very hard. If Google can auto-create these, as it does for many YouTube videos, then that makes the power of its search that much better even if the original podcast doesn’t have good meta-data. Users could jump straight to relevant section within a podcast. However this does raise questions of accuracy, and perhaps more so, intellectual property in ownership of those virtual transcripts (Cf All the arguments surrounding Google’s book-scanning initiatives). That all said, I’m unaware of anyone raising those issue with YouTube videos.

In summary then, a good first proper move by Google. They’re going to treat podcasts as essentially search assets, but using their Assistant to ensure that you keep track of what you have and haven’t listened to. However, I wouldn’t expect a significant overnight increase in the number of podcasts served. But podcasting overall continues to see steady growth, and this will undoubtedly help.

Garmin Varia RTL510

I seem to have a constant battle with rear lights on my bikes. The main problem is that I use a saddlebag on my full-size bike, and attaching a bike light to it is a seemingly simple task, but tends not to be brilliant.

If you have enough seat-post showing, then placing the light below the saddlebag in such a way that it’s still visible to traffic, is probably the preferred option. But in my case, there isn’t really enough seat-post showing.

Topeak seem to have the popular saddlebag market sewn up, and I have owned several of their models. However, in many instances, when you then hook a light through the slot made for them, they hang backwards and downwards, meaning that the light isn’t as effective. Remember, a rear light is basically only there for you to be seen!

My preferred rear lights, for compactness, have been Lezyne’s Zecto Drive range. But they suffer this problem.

My recent solution has been to change my saddlebag to a use a Topeak Wedge Sidekick saddlebag. I have the smaller of the two sizes. That’s enough for a tube, a couple of CO2 canisters, a large multi-tool, tyre levers and patches. Importantly, it’s firmer than other Topeak models, so hooking a light on the rear keeps the light pointing higher rather than lower. I’ve been happy so far.

All of which brings us to Garmin’s new Radar Light. Now why might I want a radar light? Is that strictly necessary? The answer is clearly not, but it has immediately proved itself useful.

The light fixes to your bike via a regular Garmin quarter-turn connection. The box includes mounts for a seat-post, but as mentioned above, I don’t have room to place it on a seat-post. Fortunately, creative people who design stuff to be 3D printed have got solutions for you. I bought a Varia Saddle Bag Clip via Shapeways. They 3D print things that creators have uploaded to order. It’s an extra cost, and it’d be nice if Garmin packaged one in their box, but it does the trick. Alongside the Topeak Wedge Sidekick, the light stays firmly pointed in the correct direction.

The light itself is relatively basic. There is a single led light and it has four modes – solid on, night flash mode, day flash mode and standby mode (As far as I can see, standby mode is a bit useless since it doesn’t have traffic detection). The battery is recharged via micro USB and the battery life seems decent with 6 hours in solid mode and 15 hours in day flash mode. Fine for most rides, but you’ll probably need a backup light if you do, say, the Dunwich Dynamo.

So how does it work in practice? While a standalone device is available (RTL511), it’s perhaps most useful when paired with compatible Garmin bike computer. In my case I paired it with my Garmin 1000 which was as simple as adding a new sensor. In the top right hand corner you get an indicator that there is connection, and you’re ready to go.

It works by determining larger objects that are moving at a different speed to you. When it sees one, it gives you an alert and small dots appear on the side of your Garmin bike computer (the right hand side by default). The device can determine several vehicles at once, and you’ll see a series of dots. The closer the dots get to the top of the screen, the closer they are to you. If a car passes particularly fast, the screen goes red, but if it’s slower then you get green. The unit will also beep to alert you to this traffic.

I must say that in practice, it worked very well. You do get the concessional false positive, and if a car stays behind you, matching your speed, perhaps up a slow windy hill with few overtaking opportunities, it may lose the vehicle for a while. Other cyclists tend not to show up, but in general I really like it. Note too that it obviously only detects traffic behind you and coming towards you. You shouldn’t see dots tailing off towards the bottom of the screen!

The radar has a 40 degree wide angle which covers a decent chunk of the road. It also means it continues to work going around corners for example. Garmin says that it can detect vehicles up to 140m away, and I’ve no reason to doubt that in my usage.

And when the vehicle gets very close, the blinking on your light increases in frequency to make sure that the driver has seen you!

The only real downside is the impact on battery life of your bike computer. The Edge 1000 I use has never had amazing battery life, but I got the low battery warning after a 70km ride last weekend which is a bit early. Obviously, the number of sensors you’re using will impact on that, as will things like screen brightness and me using maps (which I was). But while the light itself will probably last well, you’ll need to keep your bike computer’s battery topped up.

I’ve not tried the light in the city centre, and I understand that it can be less useful – probably too much other traffic to cope. In any case, you nearly always have cars behind you, so there’s little added value. It’s best for those places where it feels like cars sneak up on you.

Even with only a couple of rides under my belt, I’m already a fan.

For a much better and more detailed review, DC Rainmaker is obviously the place to go.

Facebook Pixel Tracking

This morning Nieman Lab had a really good piece asking whether if there was a certain amount of hypocrisy coming from certain news organisations castigating Facebook for leaking data, when at the same time they’re helping Facebook collect more data on you.

Recall yesterday, when I said that some of Facebook’s data was missing from the download, and I highlighted Facebook Pixel? Quora is a good place to go and have a look (Note: Quora itself has a Facebook cookie):

The pixel is a transparent, 1×1, unique image file that can be embedded on pages outside of Facebook (unique = 1 per advertiser account). That image file, however, sits on Facebook servers. So, each time it is loaded, it increments counters on Facebook’s side.

And each time the pixel file is being seen by a user… Facebook servers can see which browser is used, which machine and which IP address. In other words, they are able to reconstruct that signature – they know which Facebook user has seen the pixel.

In essence, the Facebook Pixel lets you then target people who visit your site when they’re back on Facebook. And of course, Facebook now knows that you’ve visited a particular site, deepening the picture they hold on you.

And Facebook also has the Facebook Audience Network, which basically extends Facebook’s advertising business beyond the bounds of the Facebook website. In particular, they’re targeting mobile sites and apps.

Using the EFF’s excellent Privacy Badger browser plugin, I looked at some of the UK and US’s biggest news websites to see which ones allow Facebook to track you. This obviously isn’t a comprehensive list, but it gives you an idea.

Sites with Facebook Cookies

  • The New York Times
  • The Washington Post
  • Forbes
  • The Daily Mail
  • The Sun
  • Metro
  • The Times
  • The Economist

NB. These are at time of writing.

It must be said that I’ve not really gone into detail about Facebook’s business model here, but it gives you an idea.

And there’s a wealth of data being collected by many companies beyond Facebook – and a multiplicity of ad tracking cookies going around. Upwards of 20 cookies on a website is not unusual. Sometimes they’re just there for analytics purposes. All the advertising networks use them, with Google and Facebook being by far the biggest networks globally.

And there can be good reasons to use tracking cookies. This very site uses Google Analytics to count the number of people who visit, for example. I’ve embedded Vimeo videos and Flickr images on this site, and they have tracking codes built into the code I copy to this site. If you comment, there are various ways you can log in, and they have tracking codes too. I’d prefer there not to be, but if I want to properly use those sites’ services then I have to play ball with them.

While everyone kind of knows that the pair of shoes they looked at over lunch, but didn’t buy, is now following them around the internet, and that must be using some kind of tracking information, I’m not sure that many of us really understand how widespread this is, and how much data is being captured about us.

[Update: In related news, Mozilla has announced a Firefox plugin that stops Facebook tracking you around the web. Useful if you’re not already using something like Privacy Badger or Ghostery.]

Is Netflix Quite As Smart As Everyone Says It Is?

That’s possibly a provocative title, but I’ve come to the conclusions that while Netflix is very good at some things, I’m not certain that its recommendation engine is entirely as linked up as you’d think it’d be.

A couple of recent cases in point.

I was really looking forward to the new Alex Garland film, Annihilation. While I was slightly disappointed it wasn’t getting a cinema release, I was very pleased that Netflix was investing in it (well, buying the rights), and making it available to its subscribers. I dutifully searched for it ahead of its 12 March release, and added it to “My List,” Netflix’s somewhat clunky system for saving things you want to watch.*

Although I believe the film was made available at midnight UK time, but I waited until Monday evening to open the Netflix app on my Nvidia Shield and settle back to watch. I thought that they’d probably have the film front and centre when I opened the app. After all, it was a big coup them getting it. Plus I’d explicitly added it to my list.

There was no sign of it. It wasn’t in trending (too early I guess), or in any of the top lists of things I might want to watch. I ended up using Search to find it. It was – but hidden.

Then over this past weekend, while I was out and about, I got some Instagram advertising for a film called Paradox with Darryl Hannah and Willie Nelson. I’d not heard of it, but clicked through and saw video for some kind of western themed film. “I might watch that,” I thought – vaguely intrigued. Netflix are obviously promoting it, I’d catch up with it at some point.

Later, with that thought having drifted out of my head, I did open Netflix again in search of something to watch. Had I spotted Paradox, I’d have at least given it a second look.

But it wasn’t there. Or more to the point, it wasn’t obviously visible. In any case, because a film I’d seen promoted precisely once, was no longer in my view, I didn’t search for it. I only remembered this at all because I saw a second Instagram ad for it earlier today.

But again, it feels like Netflix is being a bit slow and doesn’t have all its ducks lined up. It’s not that I don’t think they can do some clever stuff, but they’re not as good as they make out.

Have you heard of a Danish comedy drama called Rita? Maybe if you’re Danish, but otherwise, you might not have. Netflix never recommended it to me. It was someone on Twitter who noticed it. It’s very amusing.

I started watching a Spanish series called La Casa de Papel. It’s a series about a gang of thieves who try to rob the Spanish Mint. It starts well, but like another Spanish series I saw on BBC Four last year, the strong hook doesn’t last the course, and we end up with an interminable number of episodes where not a lot happens, and the villain is really villainous. More plot and fewer episodes please Spain. I mention this because after I’d watched a few episodes on Netflix, the series promptly changed its name. It’s now called “Money Heist,” although it wouldn’t be obvious to those like me who’d started watching it under another name entirely. I had no idea what Netflix had done!

I’m always suspicious of over-claims about how briliant someone’s algorithms for discovery are – mainly because I’ve yet to experience anything that’s really that good. Amazon is pretty bad at recommending me books I didn’t tell it about, and music recommendation engines are pretty poor in my experience – especially if you move beyond the obvious.

Maybe they work for some, but I’m underwhelmed.

* I say it’s clunky, because it’s incredibly binary, and doesn’t allow you to make lists for different things. Furthermore, when you watch something that was on the list, it doesn’t then remove that item from your list. I’m also not aware that Netflix alerts you when something that’s on your list is shortly to be removed. Another useful feature.