Google Poetics is a project to record the beautiful poetry that Google’s autocomplete feature sometimes writes.
Inspired by this recent example, I recorded a song the lyrics of which are entirely Google autocomplete suggestions.
For my own privacy, I avoid using free mail services like Gmail and instead run my own mail server (hosted at Brightbox, a very trustworthy cloud server provider here in the UK ;). I do use one or two other Google services which means I have a Google account which means I do actually have a Gmail address. I also have an Android phone and that has the Gmail app preinstalled… by this point you can surely see that I should give up any pretense of privacy, but I prefer clinging onto it’s remaining delicate threads. It provides an illusion of dignity otherwise missing from the Internet.
Anyway, so I happen to have a pretty convenient email account available to me when I’m mobile whether I want it or not. I don’t want to connect it to my own IMAP server, but I don’t want to start sending mail from a Gmail address as it will end up in people’s address books and they’ll start sending new mail there instead.
So I’ve found a compromise: I’ve configured my Gmail account to let me send email from my personal email address (and set up the associated SPF DNS records too). And I’ve configured by own mail server to forward back any *replies* to my Gmail address. So whilst I’m out and about, I can *send emails using Gmail if I need to and read any replies to that email*. But new email direct to my personal address is never seen by Google.
I use maildrop, so my filtering rules are in the mailfilter language:
if ( ( /^In-Reply-To: .*mail.gmail.com/ || /^References: .*mail.gmail.com/ ) && hasaddr("email@example.com") )
Any emails that are replies to Gmail emails and are to (or cc) my personal email, get forwarded on to Gmail (to an alias I can use to avoid any forwarding loops).
I could also check the
Message-ID header for
gmail.com to detect new incoming email that had originated from a Gmail account anyway – the logic there is that Google have already seen it, so my privacy has already been invaded; I may as well get something out of the bargain. Worth noting that replies to my replies to incoming email from Gmail will hit my reply forwarding rule anyway.
I sometimes hear conference speakers admit they only finished their slides minutes before their talk started. Or I see them the day before admitting that they still have to write the whole thing.
Whether you’re presenting at a huge national conference or a smaller local group, this is simply disrespectful to the people who take the time to be your audience.
The only reasonable explanation I have for why people do this is that they’re trying to manage expectations. Concerned it won’t go well they make as if it was a last minute job, even if it wasn’t. If it does turn out to be a disaster then, well, they didn’t try hard. If it’s a success, even better! They just threw a great talk together without any effort.
But when a speaker admits this kind of thing, all I hear is that they don’t care enough about my time to prepare and for some inexplicable reason, they’re terribly proud of themselves for it. Bragging about it almost.
Not all talks require major preparation but even the shortest and most basic require proper consideration.
Not everyone has time to prepare in-depth talks but if you know that’ll be a problem for you, perhaps you shouldn’t agree to speak in the first place.
If you do find yourself ill-prepared though, might I suggest that you don’t let your audience in on it.
Don’t be proud about wasting their time.
That just makes it worse.
I bought a Behringer B5 condenser microphone a couple of years ago to record my acoustic guitar. Add in a second dynamic mic I already owned and a little two channel USB preamp with phantom power and it sounded really nice.
Then after a few months the condenser mic started picking up some interference. It was a weird kind of rumble but with a kind of radio tuning sound, and the odd pop and click.
I tried changing channels, switched power supplies and cables but nothing helped.
Finally I came across a forum post describing a similar problem with the cause being humidity. Apparently condenser mics don’t do well in humid conditions and my office is a little damp. I’d left my mic out of its case a couple of times in these conditions and it got damp. The silica gel packet that came in the case should have been a clue.
Anyway, I popped my mic in my electric oven set at 30C, left it for 30mins and now it’s as good as new! Phew.
Another quadcopter crash today. I was just hovering in my garden and it randomly flew into a fence. All four (carbon fiber reinforced) propellers broken, two motor mounts bent and a leg snapped.
Looking at the telemetry logs (which I happen to have as I had my laptop connected at the time) it seems the GPS suddenly reported the position wrong by a couple of hundred metres and the geofencing feature kicked in and it tried to return to launch. More investigation needed though.
I think I might change the fail safe to just auto-land, rather than RTL.
I may also do a complete rebuild and see if I can take the opportunity to make my quadcopter lighter and bit more agile.
Actually, the GPS didn’t suddenly report a wrong position – and in fact, arducopter has protection against that exact kind of GPS glitch (it ignores sudden impossible increases in GPS position).
What actually happened is the GPS position slowly drifted away from where it really was, avoiding the arducopter glitch detection, until it went outside my configured geofence distance. At that point, it went into failsafe mode and tried to fly where it thought home was (RTL – return to land).
So, I’ve changed the failsafe mode from RTL to just “land” for now and am investigating the GPS problem (though I was at the bottom of my garden which has poor GPS reception).
I have a Lenovo Thinkpad T431s running Ubuntu. For those of us not running Windows, Lenovo provide a bootable CD image for their firmware updates. The Thinkpad T431s does not have a CD drive and the ISOs Lenovo provide will not boot from a USB key.
It’s trivial to provide images that can boot from a USB key, but Lenovo instead just assume you have an external USB drive because they’re dickheads; or incompetent. Or both. Let’s assume both!
Anyway, you can convert their images into something you can write to a USB key if you know the magic incantation.
You just need the geteltorito command, which is provided by the genisoimage package on Ubuntu.
The isoinfo command (in the same package) shows that the iso has an ElTorito virtual disk section, which you can’t see normally – it’ll just look like an empty CD if you mount it.
isoinfo -i fwsx04.iso -d CD-ROM is in ISO 9660 format System id: Volume id: FWSX04 Volume set id: Publisher id: Data preparer id: Application id: NERO BURNING ROM Copyright File id: Abstract File id: Bibliographic File id: Volume set size is: 1 Volume set sequence number is: 1 Logical block size is: 2048 Volume size is: 15542 El Torito VD version 1 found, boot catalog is in sector 20 Joliet with UCS level 3 found NO Rock Ridge present Eltorito validation header: Hid 1 Arch 0 (x86) ID 'NERO BURNING ROM' Key 55 AA Eltorito defaultboot header: Bootid 88 (bootable) Boot media 4 (Hard Disk Emulation) Load segment 7C0 Sys type 6 Nsect 1 Bootoff 1B 27
And you can extract the ElTorito image like this:
geteltorito fwsx04.iso > fwsx04.img
And then you can just write that image file direct to your USB key. In my case, my USB key shows up as /dev/sdb
dd if=fwsx04.img of=/dev/sdb
then you can shutdown and boot from the USB key and the firmware flashing utility will execute.
Or you could buy a Lenovo external DVD drive just for doing firmware updates; yeah thanks Lenovo.
I crashed my quadcopter last week due to flying out of range of my transmitter, causing my quadcopter to drop out of the sky. This happened because I hadn’t calibrated my throttle failsafe properly.
This is for my Spektrum DX8 transmitter and AR8000 receiver, but it’ll be the same for most Spektrum devices (and the theory is most likely the same for any transmitter/receiver)
The theory is simple: Your receiver records your transmitter’s throttle position when you first bind them together. From then on, if the receiver loses contact with the transmitter, it defaults to outputting that throttle setting to your flight controller.
So usually, you’d make sure your throttle is down when you bind. So then, if you’re flying along at high throttle and go out of range, the receiver will just throttle down (and your quadcopter will drop out of the sky like a potato without a parachute).
Most flight controllers (mine is an APM 2.5 running ArduCopter 3.1.1) are able to land (or return to the launch point) automatically, but it needs to be able to differentiate between losing signal and you just throttling down (maybe you’re doing a stunt, or testing Newton’s law of universal gravitation).
The trick is, to have your transmitter output a lower throttle setting than the lowest stick position when you bind it to the receiver. That will allow your flight controller to tell the difference between you throttling down, and your receiver going into failsafe mode.
There are a couple of ways to get your DX8 transmitter to go lower than the lowest throttle tick position but the simplest I found was setting the throttle travel setting temporarily:
Now, your lowest throttle stick position is outputting a value of something like 920 to your flight controller. Now you go through the binding procedure. As a reminder:
Now you go back into the “Server Setup” menu on your transmitter and reset the travel back down to 100%, so your lowest throttle stick position is back to normal (around 1100).
So now, your lowest throttle stick position outputs something like 1100 to your flight controller, but if you power off your transmitter it will drop to something around 920.
Now you can configure your flight controller to land, RTL or quantum leap or whatever when it can detect that it is out of range. On my ArduCopter, the defaults were all good (anything below 975 detected as failsafe) – I just had to enable failsafe.
I was listening to a Philosophy Bites interview with Gary Francione about animal abolitionism. Animal abolitionists argue that animal welfare reform is nonsense and animals should not be regarded as things to be owned and used. He’s very compelling and his apparently clear reasoning is quite convincing.
During the podcast Francione explained the philosophy of Peter Singer, considered the founder of the animal rights movement. Singer essentially argues animals don’t understand death therefore they don’t suffer knowing it’s coming so killing them is not causing suffering.
Francione doesn’t agree and counters by arguing that the animal would obviously prefer to live rather than be killed.
At a first look, this seems to be clear reasoning but in actual fact it’s a bit of a switcheroo. Singer says it is ok to kill animals because they cannot reason about their own death. Francione says that if they could reason about their own death, they’d prefer to live, so killing them is morally wrong. But Singer’s argument would not apply to an animal that could reason about its own death!
By instilling the animal with the power of reason, Francione transforms it into an animal that Singer would not support the killing of anyway. Singer’s argument only applies to animals that cannot reason in that way!
It’s important for me to make clear that I think Francione’s argument for veganism is probably stronger than Singer’s argument for eating meat (or at least stronger than my own understanding of Singer’s argument at this stage). Beyond the theory, it’s not really possible to avoid unnecessary suffering to an animal during the meat production process. And any suffering at all is arguably unnecessary, since most people do not need to eat meat to survive. It’s not quite this simple, but I think Francione is broadly right.
I still eat meat though. I’m generally quite careful about the source of my meat, and often eat vegetarian wherever I can’t be sure of the source but I’m still not really convinced I should be eating meat.
Yet I still do eat meat. It’s easy to say humans are rational beings and how that separates us from other animals, but how many of us actually live rationally? Clearly I don’t!
I had a major quadcopter crash a few weeks ago – and it wasn’t due to my dodgy flying skills. One of my propellers broke in mid-air; one blade snapped off from a prop at the hub. The prop was one of two new bright orange coloured ones I’d bought to help keep track of the direction of the aircraft, which worked very well for me until it crashed. Looking more closely at these new orange props I now notice that they are much thinner where they attach to the hub. My other props thicken out as they approach the hub making them much stronger. The orange ones are much weaker – bending them quickly results in white crease lines at the hub, which doesn’t happen with my black ones. Lesson learned, don’t buy cheap props (actually, they weren’t any cheaper, so the lesson is to understand how to recognise crap props).
At the time the prop snapped, I was about 20 metres up in the air, traveling horizontally at quite some speed. When it broke, which made a “ping” sound, I panicked and dropped the throttle and then realised that was wrong and throttled back up to try and cushion the landing. I’m not really sure to what extent that helped, but it hit the grassy ground hard. I was flying it in a field near my house and was prevented from immediately retrieving it by a group of curious horses that went to investigate this UFO crash landing site. Luckily they decided against trampling or eating it and eventually lost interest.
It was still powered up when I got to it, though unarmed. Annoyingly I’d decided against flying with the camera this time so didn’t have any dramatic footage of the accident.
The battery looks too badly damaged to use. It’s not quite punctured, but it’s bashed up and misshapen. I may try recharging it out of curiosity, but I expect it to set on fire, explode or quantum leap or something – who knows with these lipo things. Quantum lipo. A friend has loaned me two new batteries, which are higher capacity (3900mAh) but weigh more – so we’ll see how that works out for performance and flight time.
A couple of the other props got damaged, but the motors, ESCs, and controller stuff seem all fine.
The two arms where it hit the ground were bent quite badly.
It’s taken me about a month to get around to it, but I’ve finally rebuilt it. The new frame is pretty much the same design as before but with a couple of minor modifications. I mounted the arms less closely to the middle of the base, so there is more space for cables, and I mounted the motors about 1cm closer to the end of the arms. So the result is that the motors are about 2-3cm further away from the base, which I expect to change the behaviour slightly and gives me more headroom between the props and the base – they were quite close before and prevented me from mounting some bits close to the edge of the base.
Since my first build I’ve acquired a power drill mounting and I built a wooden jig to keep the two bases in place, which meant I was able to drill holes much more accurately. I also drilled lots of extra holes which I knew would be handy, so more places to run cables and pass through tie-clips. It’s much tidier now.
Lastly, in an effort to isolate vibration better, I put patches of neoprene where the arms are bolted to the base plates, and where the motors are bolted to the arms. The neoprene was an old laptop “skin” I cut up. I had to remove it from the motors as it meant they weren’t mounted perfectly square, so the craft was spinning a bit. I’m not really certain how effective this is as I don’t really have any data to back it up so it’s a bit cargo culting atm. I may experiment with this more scientifically at some point.
I bought some much sturdier carbon-fibre reinforced propellers too – I’ll write about my efforts to balance them properly later.
In summary, crashed it due to cheap propellers and I took the opportunity to rebuild the frame with some improvements. No horses were harmed.
The Linux kernel exposes lots of interesting information via the
/proc filesystem. For example,
/proc/net/udp expose information about all tcp and udp sessions on the server.
In the usual Linux kernel style, these files are freely readable by all users on the Linux system. This becomes a bit of a problem on a multi-tenant system like Heroku. Heroku deploy many customers apps on each of their Amazon EC2 servers. So if you create an app skeleton, deploy to Heroku and shell out and you can see all the other apps tcp and udp sessions:
$ heroku run console
Running `console` attached to terminal... up, run.1971
irb(main):001:0> system("netstat | grep ESTAB | head -n 10")
tcp 0 0 22d87d80-deb7-415:33113 ip-10-40-86-97.ec:27018 ESTABLISHED
tcp 0 0 22d87d80-deb7-415:57256 ip-10-60-122:postgresql ESTABLISHED
tcp 0 0 22d87d80-deb7-415:59268 domU-12-31-3:postgresql ESTABLISHED
tcp 0 0 22d87d80-deb7-415:38536 collector6.newrelic:www ESTABLISHED
tcp 0 0 22d87d80-deb7-415:54459 ip-10-189-243-92.:27017 ESTABLISHED
tcp 0 0 22d87d80-deb7-415:50495 ip-10-151-25-11.ec:6242 ESTABLISHED
tcp 0 0 22d87d80-deb7-415:53192 ip-10-114-25:postgresql ESTABLISHED
tcp 0 0 22d87d80-deb7-415:47405 22.214.171.124:https ESTABLISHED
tcp 0 0 22d87d80-deb7-415:39011 collector6.newrelic:www ESTABLISHED
tcp 0 0 22d87d80-deb7-415:52237 ip-10-218-31-147.:42002 ESTABLISHED
If the netstat command (or system method) is not available, you can just read and parse the file in Ruby.
It’s an information leak, so I think it is quite low risk but still a hint that Heroku’s process separation still isn’t as strict as one might hope (I still remember David Chen’s pretty shocking discovery back in 2011)
There are a few ways to fix this kind of problem. Mandatory access control systems, like AppArmor can prevent processes reading these files. The Grsecurity security patches have lots of protections against proc based information leaking too, these files included.
I reported my findings to Heroku’s security team back in September 2012. I had some problems getting timely responses at first but after whining on Twitter (in March 2013) I got lots of attention from them (and an apology for lack of response). Complaining on Twitter is a clearly a powerful tool and to be used only wisely; and maybe sometimes when drunk.
Anyway, it seems they had been working hard on this and just failing to update me; it is quite a major change for them and a low priority bug imo. They’ve fixed it by properly virtualising networking too. It’s mentioned on their change log but it doesn’t go into too much detail. Important point is that it’s fixed.
It turns out that whilst I have the ability to build a quadcopter (it’s actually not that hard) I don’t quite have the ability to fly the thing around without risking crashing it into nearby fleshy things, or trees.
And whilst there is a little glee involved in a crash (it gives you things to fix, yay!) it slows the learning process somewhat and the propeller smashing can be expensive.
I tried to get the flightgear simulator working so I could practice on my computer, but it’s really complicated and I had little success.
So I wrote a game to help me practice. It’s a 2D top-down quadcopter simulator. It makes no attempt to seriously simulate real physics or even a real a quadcopter. Instead, it aims to just help you become familiar with the transposing of controls as the aircraft rotates around. It’s meant to be controlled using your real radio control, via an ArduPilot controller.
You can see it in action here:
It doesn’t really have a goal at the moment, you just fly around the screen. Once it a while it randomly perturbs the aircraft, sending it spinning off in some direction as if it clipped the ground or something and you have to rectify it. The perturbing is interesting, because I do fine flying around slowly but panic when I clip something.
I was very bad when I first started playing, but I’ve been getting better and it’s definitely translated to better real-world flying.
I wrote it in Ruby, using the Gosu game library (which has a few native dependencies). I’ve only tested it on Linux (Ubuntu). The code is on github. Works for me.
I’ve been doing a few test flights of my Arducopter-based quadcopter over the last couple of weeks, when it wasn’t snowing. The first flight test wasn’t so successful. Whilst it flew, I couldn’t get it much above about 1-2 metres off the ground even at full throttle. This turned out to be due to miscalibrated ESCs, which didn’t know what a full throttle signal from my transmitter looked like.
Once recalibrated, it flew pretty well but my inexperience led to a few crash landings and a few broken propellers (no major damage though). It also drifted around quite a bit, but I improved that by recalibrating the accelerometers much more carefully.
I tested out the “return to launch” (RTL) feature a few times too. The idea is that the quadcopter remembers the gps coordinates of the launch point, and when RTL mode is enabled, it flies home automatically. The exact behaviour involves it flying up to a safe altitude (above any possible trees and such), then flies home, holds it’s position there for 5 seconds and then lands.
The first time I tested this, it seemed to work but instead of holding its position, it flew off into a nearby tree. I tested it a couple more times and it worked once, but veered off again another time. I later discovered reported problems with gps accuracy with the Arducopter firmware version I was using (2.9.1). Upgrading to 2.9.1b seems to have improved the behaviour a lot, but RTL isn’t yet something I’m comfortable trusting.
I did discover that the arducopter firmware keeps quite detailed telemetry logs of its flights, along with gps coordinates and altitude. Very good for figuring out what went wrong afterwards.
Control-wise, a major breakthrough for me was realising I didn’t necessarily need to keep pointing the “front” of the aircraft where I wanted to fly. It can go in any direction at any time, you just need to keep track of where the front is pointing to know where roll and pitch will get you. This is admittedly a bit hard to keep track of, so I think I’ll paint the legs or something to help with that. When it’s up high, especially with the light behind it, it’s quite hard to tell whether it is turning towards or away from you. Need more practise.
I also had some problems with the legs sticking in the ground (or breaking) on heavy landings. I’m making some improvements and might blog about that later.
And I stuck an old plastic takeaway container on top to protect a bit from rain and snow. Which worked well until I crash landed it today and smashed it.
And today, I strapped an old Android smartphone to it and recorded a first person view of a flight. It was an old cheap phone so the video quality is poor, but it’s still exciting. I’m researching what cameras I can get to record higher definition video atm – I’ll post about that later.
Overall, it’s been quite stable and much sturdier than I’d imagined. It’s a little stressful to fly but I’ve managed to resist the urge to panic and drop throttle any time I’m not certain what is going on. I’d definitely recommend doing your test flights in a large empty grassy field (not a hard road surface), well away from people and buying quite a few spare propellers (which I did).
I’ve gathered all my videos so far into this Youtube playlist.