 Hello. Thank you. Cool, you know me. So yes, as I said, my name's Sam Mitchin. I do interesting stuff with phones mostly. For my sins, I've also been managing the phone operations here, so the various things we had like decked and bits and pieces of what's at desk phones and some of the stuff that Matthew talked about with connecting the pots phones. I've overseen the phone network, but I also get involved with doing things with artists that want to do interesting creative technology stuff through various connections. And this is a story about a project I got involved in, trying to get the data, about four years ago, four or five years ago now, and it's sort of the journey it took and what went wrong and what we discovered and things. So the project itself is called Pick Me Up and Hold Me Tight, and it was a creative piece. I'll come on to that a bit more, but just as I said, the content warning, the project itself was brought about to raise awareness of male suicide. This talk is predominantly about the implementation, so we're not really going too much into the content. There is a video we might play which talks a bit about the creative side and there's an opportunity for you to try it. If you've got a decked phone with you, keep it out because at the end we'll try a live demo which will probably break everything because that's always fun. But, yeah, this is predominantly a technical or a narrative talk of building technology. So, yeah, hence the war dialling the payphone estate. So, yeah, how it started. Back in summer of 2018, I discovered this project called Pick Me Up and Hold Me Tight for some friends. It was launching a crowd funder and the aim of the project was, as he said, on the 1st of January, they didn't say which year, at 11am we are going to make every payphone in the UK ring at the same time as an art piece to raise awareness of male suicide. Apparently 1st of January, New Year's Day, 11am is one of the peak times for that kind of thing. So, it was, and payphones, there's a lot of connections to loneliness and there's lots of people call via payphones every payphone you go into and they still keep payphones in certain locations which might be prominent for that. I just thought it sounded like a really interesting project. It was a good worthy cause and I was kind of interested in the tech side of it as to, wow, okay, making all the payphones ring at once. That's not easy. So, I put 20 quid into the crowd funder via a friend and he sort of said, oh, that sounds really interesting. You know, I'd love to know how they're going to do it. And I got an email back from the guy behind it saying, Clare says you know stuff about phones. And I'm like, yeah, yeah, I'd love to know what your plans are. And he's like, do you know how we can make all the payphones ring? So, I thought I was backing a project to find out how they were going to do it and I turned out that I was the one that was going to do it. Which is kind of the story of my life really. So, yeah, they started off with a crowd funder on, it wasn't Kickstarter but one of these, kind of more of an arts platform, which one. And initially, BT payphones, the payphone division of BT, excuse me, were kind of on board as a supporter. It was a project that they were like, yes, we'll help where we can as long as it doesn't require them doing any work or giving you any money, basically. So really, they had a creative vision. So George and Jada, the two creative directors behind the project had an idea as to what they wanted the creative piece to look like but no real idea how they were going to deliver this. I think they'd done a few little things with small payphones with a couple of hacks on Twilio. At the time, BT's UK payphones list, so we'll go into the different types of payphones later, but there was 34,000 payphones in the UK on their list as of summer 2018. And we actually had an Excel file with the numbers, locations of every payphone at the time that were BT public payphones. So what that kind of is, is the ones basically on the street. So there are certain ones that you might think are BT payphones. Obviously the ones in pubs and things that don't look like, you know, that are just run by the pub might not be, but also all the ones in railway stations didn't count, all the ones in shopping centres, anything kind of on private property was a separate subset, but we had all the public ones, which was, you know, a pretty good definition of payphone. So, yeah, we started, I started figuring out how I was going to make this thing work. So lots of conversations with the artists, what do they actually want this kind of experience to be? What do they, you know, how do they, what happens? Okay, I make all the payphones ring, then what? And so they had a, at the time, interactive audio piece. So it was a lot around kind of listening and, you know, understanding people. And it guided you through asking you questions around how good a listener do you think you are, and asking you to, like, press one to five to answer this kind of survey thing. And then at the end it would do some maths on looking at the answers you'd put in and give you a sort of feedback on your listening skills. Which, you know, actually relatively easy, anybody that's done phone development, okay, you want to make an outbound call and you want to drop that into an IVR and collect inputs and place in audio files and then do a little bit of logic with the inputs at the end. So I started prototyping stuff. I was working for Nexmo at the time, so using, you know, any of those kind of platforms are quite easy for building this stuff and we worked through a lot of the creative stuff. And at the same time talking to, or designing the platform as to how am I going to do that many calls at once, so to speak. And there was a little bit of discussion around, okay, when you say 11 a.m. you want all the phones to ring, do you want, like, literally click, tick over 11 a.m. every single payphone is in sync or, you know, how much leeway do we really have here? Is within 15 minutes fine, you know, if they all ring sometime from 11, between 11 and 11.15, and we had lots of back and forward. So finding that kind of balancing act as to what's really feasible, because actually to try and dump 34,000 calls in sync was probably not going to be particularly easy. And had some conversations with BT, because we had BT at the time as a partner, about, hey, can I send you 34,000 calls? You know, where do you want me to send them? And they were a little bit kind of, that's a lot of traffic at once. Can you do 500 a second? But that was, yeah, the limiting side of any of their platforms. They didn't have any sort of integration points. But, you know, we were progressing. So the plan was to do a pilot in Liverpool, because one of the creative institutions that was involved was the Liverpool, got me down, Liverpool Arts Institute or something, on the 1st of January 2019. So we were actually going to use just a third-party SIP provider. The number of phones in Liverpool was a couple of hundred. We did an enterprise for our list, so it was a small-scale kind of pilot. And then on the 18th of December, so 12 days before we're due to go, and a week before Christmas, BT pulled out. They said, no, we're not kind of, we're not really happy with the project, we've decided not. Not so much from a technical standpoint, and the traffic, because we were just, I was sorting that. Actually what they were more concerned about was basically anything drawing attention to pay phones, because they were all in a pretty poor state, the amount that were broken, not working, vandalised, generally filthy, disgusting condition. And pay phones, it turns out, within BT are obviously a bit of a... They're not exactly the glamorous end of the business, are you? You're not doing 5G and fibre broadband if you're the head of pay phones. So, yeah, they pulled out. So we kind of had to just, a week before Christmas it was like, okay, this is not kind of feasible to kind of, we have to rethink a bit of this. So we put a hold on it. Early 2019, we got back and sort of said, okay, can we do this without BT? So at the time we were using BT's list of data, one of the key things they gave us was this list of pay phones. And if they've sort of pulled out of the project, is that really our data? Can we still legitimately use that? So we had lots of conversations with lawyers, and the output of it was that we decided that we would source alternative pay phone data. So there are various lists floating around the internet of lists of pay phone numbers in the UK. So between that and some human intervention, we distilled down to a unique list and we kind of knew what sort of numbers we were aiming for. So we had a reasonable idea. So we created our own list of pay phone data and we spread the calls out. So actually we'll do it over 30 minutes. We'll scale that back a bit. With a plan to do our pilot in Leeds first of Jan 2020. So we basically delayed a year. So we moved to Leeds and we actually sent some people out. We recruited a whole bunch of volunteers. I think we had 60, 70 volunteers who went round with our list of pay phones in Leeds and audited them and basically said, right, here's a list of pay phones. Can you go and check if they're still there? If the number is what they think it is? All kinds of things. See if you find any others. There are photos which was really handy because it was like, send us back a photo of each one so we can get an idea. So these are just some of the photos that got dumped in. You can see down the bottom there, particularly these ones with the screens, these are these new inlink kiosks that are sort of appearing, which are a whole lot of talking themselves. But basically they're an advertising board but by putting a pay phone in it they can avoid the need for planning permission to actually put up an advertising horde. So they put a tiny little phone on the edge of the thing and put an advertising board on it but BT can install pay phones on the public highway without needing planning permission or paying rent to the council or anything. So they use that as a loophole to sell advertising. And actually the calls on them are free because it's not worth charging for that. They make more money off the ads on the billboard and they don't have to put a sensor around to collect the 50Ps out the box. So, you know, we were kind of on target and then we hit some technical difficulties. So somewhere around the summer of 2019, so we did before this was, we were doing this survey thing. You know, you called, you picked up and then it asked you to press one to five based on your answers. BT made a change to their pay phones that prevented DTMF, so the keypad working, if it was an incoming call. I still don't really know why they did this, necessarily how. I mean, you know, pay phones all have a digital kind of, they call back for a control. So, but they, I don't think it was kind of to scupper us. I think this was just somewhere along the line, somebody rolled out some new software and nobody really ever considered this use case. Because actually nothing ever, you know, one pay phones very rarely receive incoming calls and pay phones almost never receive incoming calls from automated systems that then ask you to navigate a menu. So we were kind of an edge case here that we caught up on, but somewhere along the line they changed that. So that kind of, you know, gave us another spanner in the works. So we had to rewrite, really this was more a problem. I mean, it was a problem for me for the first few months because we started just getting initial reports on these tests going out in Leeds of like, I called, I answered, I picked up but it wouldn't do anything and I'm like, that's weird and I was trying it and I was like, well the IVR stuff works fine on my phone. I went to like the pay phone in my village, still worked. And, you know, we're trying to track it down. That's why we started getting a lot of the photos of the pay phones. It's like, well send me a photo of the phone it works on. I was trying to work out if there were certain models maybe it didn't work on or were they using like some non-BT ones or something. But it just seems to be that the software was rolling out and turns out the pay phone in my village that I was using to go and test this on hadn't been updated and was still working and the ones in Leeds got updated sooner. I think actually they were the top of BT's list because it went in area code numbering order and O113 is one of the lowest numerical order area codes. So, yeah, we had some fun with that. Once we got to the bottom of it and I kind of actually managed to find some phones that it did fail on, we rewrote the audio so it's now just a single plays back a nine minute I think it is audio file, which made the tech a bit easier for me but so I wasn't collecting inputs. So at this point now we had a working platform. So I built a scalable calling platform a system basically to make a lot of phone calls in a relatively short period of time and we were going to use it for a one off. We're going to fire this up in New Year's morning at about nine o'clock and we're going to be done with it by midday. So hence this sounds like an ideal thing for cloud computing. We need a lot of capacity for a short period of time but we don't want to pay for it all ahead of that. So I used AWS FreeSwitch if you're familiar with that was what we used as our calling platform so the platforms you have like Twilio, Nexmo these kind of things, the programmable communications these are great but they have capacity limits. I think Twilio limit you to one call per second on a standard plan. Nexmo I think is 30 simultaneous calls. They all have some kind of limit like that so that wasn't really going to scale for our needs. FreeSwitch and just SIP providers who give us a phone trunk would give us a lot more flexibility in how we do that and then a little bit of serverless kind of back end control. So we get into the slightly more technical part of the talk with a very badly drawn AWS diagram. Apologies if anybody works for AWS I'm sure this breaches a whole bunch of the guidelines. This was kind of how it all came together. So I keep looking for a screen behind me there. Over on the right this was our sort of ingestion so there was me and a list of CSV files and some Python scripts basically de-duping CSV files and trying to get this single set of payphone data into DynamoDB. Dynamo is where we held the records where we one record per phone some geocoded data so there was effectively a lat long that we took from post codes we mostly had of them the number, a description and just an ID that we used internally. They all went into DynamoDB and then we had a Lambda function which was actually what would kick off the call so select all from this table on the database and drop them into simple queue service into an SQS queue so when we want to actually make the call I just invoke the Lambda trigger from a command line and we dumped 30,000 records into SQS then we have the kind of the meter this is down the bottom there we had a whole bunch of EC2 servers there's six there I think at one point we had 14 in the kind of cluster or not really a cluster because they were individual but all running free switch with this audio file on it and the kind of the simple dial plan that says call the number when it picks up play the audio file and a little bit of Python that was just binding it to the SQS queue free switch bot code was doing was look sitting on the queue picking a number off the queue and giving it to free switch and saying place a call and each free switch box had a limit on the number of simultaneous calls it would handle so it would basically start dialing numbers until it had however many active calls that box would handle at a time I think about 100 calls per box and then when a call ended either because it had wrung out and not been answered or it had been completed or it had failed in an error it would just pick another number off so we basically had a big pool of workers pulling numbers off the queue making phone calls and then we distributed that across various SIP providers so again most of them will quite easily with a credit card and things give you 30 channels or something per account but if you want more they want like a year's commitment and you have to talk to somebody in sales and probably go out for a steak dinner and all that kind of stuff that comes with telecoms so we just set up a bunch of different SIP accounts and we didn't really care about the numbers or where it was coming from pay phones don't really have caller ID so we farmed the calls out across lots and lots of different SIP providers as a 14 basically at the most or 14 accounts a couple of them we managed to double up on all of those converged back on BT obviously the one nice thing about spreading it across lots of providers is that most of the providers are interconnected with BT at different points so we weren't sending all the calls into BT's network on like one little local exchange somewhere that was going to get utterly swamped we were calling across the country we were distributing our traffic across the country the most central thing was probably EU West 1 in Dublin or something so all the calls onto BT and then out to the pay phones the other piece I should mention that we did was as part of the credit there was a there is still a website Google picked me up and hold me tight or something it's zoo UK is the the arts group behind it they wanted to have a live map showing when the phones were ringing, when they were being answered all of this kind of thing like an interactive visual piece to because actually you need something to show just an audience as well as the participants so this was a relatively standard bit of web app stuff basically the free switch box when the call events, when the call was answered or when the call failed or whatever was publishing those publishing those events back to SQS we had a lambda picking up off that queue and updating the status in the database so again DynamoDB was being updated as to whether the phone had been answered or not and we published those out over Pusha the web socket service and then there's a web page static web page bit of JavaScript hosted on an S3 bucket which when you load the page it used another call, the historical what's the current state of the phones that have all been answered out of the database as a big jason object and plot push pins on google maps so you can see them, I'll show you a better view of the map but there's a little icon for every phone on a google map and there's a little cluster a phone box icon for when there's a clustering of them together if you zoom out so it would update them and then Pusha would actually update the page as phones were picked up so when phones were ringing they'd start bouncing when they were answered they'd go red I think it was it changed the colours and things like that so it was just a bit of an interactive thing so that's a 34,000 points on a google map with a bit of real time updating isn't particularly particularly interesting so yeah we ran our leads pilot it was mostly successful the really big thing here is now we're two but two and a bit years into the project no 18 months sorry into the project last few years I've just merged into one but we were the thing that was really noticeable was the rate of pay phones declining so when we started this project there was 34,000 on the current BT list we were down to I think at this point we reckon there was probably about 17,000 still active in the UK and falling so every time our data was going out of date and mostly the numbers were just disconnected we were getting an error code back of various descriptions depending on the provider that was looking at that phones being terminated that line no longer exists the phone box has been ripped out and sure enough occasionally you could check on things like google street view I spent a lot of time looking for pay phones on google street view and saying well we've got to go back in time it was there then it was there then it was there no it's not there anymore so I was doing all of this from home but I wasn't having to drive around but the number of pay phones was really falling so then it was a case of prepping for the nationwide event so this was just mostly refining our data we ran some brief tests so overnight at five in the morning I was calling a whole selection of pay phones every night and seeing which ones were still giving me ringing just waiting for a ringing and hanging up so it's tidy short calls that weren't connecting spread out so we weren't kind of noticeable as through any kind of error reporting or something on BT and just trying to kind of keep on top of cleaning that data up and making sure it was still relevant and that takes us to 1st Jan 2021 so as you'll probably also notice the whole Covid thing and the middle of this as well didn't particularly help fortunately I can't believe I think we were in whichever tier it was at the time that meant that with some discussion with lawyers and things it was like well if people can go out for a walk if they're going to a pay phone by themselves this is perfectly acceptable so we did our nationwide event we had at that point 17,000 numbers in our database we placed that should be 10,000 calls places a typo a microslides there so 17,000 numbers in the database when we actually made the calls 10,000 calls came back with a successful ringing kind of message and 900 of them were answered and that's over I think about half an hour so between 11am and 11.30 on New Year's Day 900 people picked or the phones were answered not all of them went and there was a reason why it was quite a long audio piece but you know some people did listen all the way through and there was obviously some publicity around this and kind of promotion in different areas and campaigns to encourage people to go and take part and then we got invited to the Compass Festival in Leeds which was this should have been 2020 this was delayed into 2021 to basically run the project again just on the phones in Leeds but every day for two weeks and then there is a short video which are we doing for time we'll try the video sounds up so this is about the project in Leeds kind of the project in general the project came about five years ago because we had lost a lot of people who had decided to take themselves away from the world and that hurt because I myself have suffered with suicidation you know in some ways I felt like I was actually living proof that there can be a way out of that darkness it's loud in the world these days all of these friends and colleagues and loved ones were men the statistics in this area suggested that the lead on this was older men whereas previously had been younger men and at that time we were working a lot with telecommunication technology and we thought it would be really incredible to be able to ring all the phone booths because they are just essentially very lonely creatures and the project just grew from making all the phones ring to actually visiting each and every phone across the whole of Leeds and actually mapping them and this was a project that took volunteers a huge community of Leeds residents that just signed up for going out and mapping the phones and testing each and every one of them it's very much a live map and it's a snapshot of this moment we're thinking about removing this payphone our research shows that this payphone just isn't used enough for us to carry on running it that some areas are perhaps less heard than others Phone boxes to me chimed really well with this feeling of you know like obsoleteness unused, forgotten like losing that vital service that they once provided and I felt like that chimed well with this new demographic that we were seeing because of it so that's the creative kind of trade for it and what I want to try now a couple of minutes is one risky live demo which may break the decked infrastructure in here as well because I don't think we've put enough antennas in but we'll find out how many calls so if you've got a decked phone if you dial that number briefly it'll just say thank you for joining and hang up on you that's putting you all into a queue when I've done that I'll then initiate something which should make everybody's phone ring and when you answer you'll hear the audio piece give it a go, I'll give you a minute we've got about a minute so I'll do that and then we'll see if we can break something you're getting messages saying thank you something, yeah has everybody done that off the decked phone what's it phone anybody, I can't really see but wave if you're still trying to get it to work or something so 5559001 on decked it has to be on the EMF network you can still hear a few clicks okay that's about a minute let's see what this does so if I now call another number okay we'll see how many ring and then how many can answer it's probably our limited capacity now you should be hearing the audio piece cool if you want to try it later if you do a different number 9002 that'll just let you dial in and listen to the whole audio piece if you don't like one of the phone booths we have on the site or something it's kind of quite emotive or anything like that so I'll be around outside if you want to ask me questions at the end thank you very much