On Maps. With Michal Migurski.

Hi, everyone. Data stories number 20. Hi, Moritz, how are you?

Hi, Enrico, how are you doing?

Good, good. So we've been doing this thing for long enough that now it comes very natural for me to say, hi, everyone. Number something.

You had some time to practice.

Yeah, we should actually change it at some point. Or maybe keep it. Yeah, that's the brand, right?

Exactly.

What's going on?

I'm doing super well. I just came back from Belgrade. I was there for resonate festival.

Oh, fantastic.

Which was crazy. Yeah, I was only there 36 hours. I had two cancelled flights. So, yeah, it was a bit of a pain. But these 36 hours, they were like a week. Super much input.

And Max, super dance.

Yeah, I was dancing too. The people who were around on Saturday evening can attest that I was dancing.

So it was some techno or what?

There was electro from Berlin Dinky.

Wow.

Yeah, cool. And I think also Belgradian DJ team. So, yeah, that's good.

Yeah. That's when you say to your wife, I'm traveling for work.

Yeah, super busy.

Super busy. Right. And then there is a movie of you jumping around somewhere in YouTube. Right?

I hope she doesn't check YouTube.

Exactly.

Yeah, no, and the conference was really good, too, so it was more in the direction of creative code, generative design.

Any highlight from the conference?

Really good. There was lots of good stuff. I saw memo Acton for the first time. Who's great? Casey Reas was there. Lots of great people. How are you doing? What are you up to?

Working. Working hard. Nothing special right now.

Semester is slowly zoning in.

Yeah, yeah. I was so depressed last week because. No, actually this week as well, because it's. No, last week it was Springdez break. But I've been working like. Like hell. Anyway, so I met my students yesterday and they asked, how was your break? And then I reminded. I said, what break? I had even forgotten about this break. Yeah.

Anyway, in your break you managed to catch up with the email of your main achievement there.

Yeah. And the inbox is growing and growing. But it's okay. Sometimes it's like dead end. I know. It's not gonna last forever. It's gonna be better soon.

Anyways, we have a guest today.

Yeah, sure. Let's introduce our guest.

So we have Mike Migurski. Hi, Mike.

Hello.

Hi, Mike.

How's it going?

Good, how are you?

Great. Great.

Cool. Cool, cool. So we have a guest and a topic today, which is totally cool. And the topic is maps. Because Mike is really one of the super map super experts, I would say. Or at least he's been doing that thing for a while. For those who don't know him, Mike has been a longtime partner at Stamen probably almost ten years, something like this.

Cool. Cool, cool. So we have a guest and a topic today, which is totally cool. And the topic is maps. Because Mike is really one of the super map super experts, I would say. Or at least he's been doing that thing for a while. For those who don't know him, Mike has been a longtime partner at Stamen probably almost ten years, something like this.

Almost ten years, yeah.

Crazy. And of course, stamen has always been known, I guess, primarily for the really advanced mapping, mapping projects and how they rethought maps. And Mike's been a big part of that. And so, yeah, we thought it would be great to have him here and talk about maps a bit and how it all changed in the last ten years. I think cartography has changed the last ten years, probably more than the last thousand years, almost.

Yeah. There's been a good two or three solid, huge shifts over the time. And for me personally as well, I just left stamen a few months ago. This is sort of an exciting time for me as well.

Right, right. Yeah. To new things. And, yeah, I'm sure there's a lot of things to explore in that area. So what we thought is we might go through the whole from end to beginning in 40 minutes or so. It sounds achievable, right? I mean, there's no problem there. So, Mike, if you think back, how did you get started in the whole, how did you get started with computers anyways?

Right, right. Yeah. To new things. And, yeah, I'm sure there's a lot of things to explore in that area. So what we thought is we might go through the whole from end to beginning in 40 minutes or so. It sounds achievable, right? I mean, there's no problem there. So, Mike, if you think back, how did you get started in the whole, how did you get started with computers anyways?

With computers? That's fun. We never asked that.

Well, I had a Macintosh plus with two megabytes of ram that my father bought.

Excellent.

In 1986.

Yeah, yeah, yeah. That's really early start. Wow.

Yeah. And it was, it was, you know, Mac draw, Mac paint. And then later, hypercard is where I got my start programming.

So you started programming with hypercard or was it basic? Yeah, I started with basic on an Atari, so.

Oh, geez. Yeah, I started with Hypercard and then basic at school shortly afterwards.

Nice, nice. Hypercard. Ah, fantastic. I never actually used it. I just know that it's legendary status, sort of.

Oh, it's amazing. And actually, Bill Atkinson, who's one of the creators of Hypercard, is speaking at IO festival this year.

I know, I know. Jeremy happened, so we have to see that, right?

Yeah, yeah. I'm gonna ask him to sign my t shirt or something.

Absolutely. Sign a hypercard for me.

But, yeah, but then throughout high school and college, I did a lot of graphic design, visual design. I was very, very heavily involved in the SF raves mailing list and the local party scene. So I was doing a lot of live video production, live projection video stuff, flyer design, things like that. And then after college, I got into code and programming. A few years later, met Eric Rodenbeck, and we started doing stamen stuff together.

And the vjing things, was that like mid and nineties or what timeframe are we looking there?

Yeah, 97 through about 99 or 2000.

Ah, fantastic.

Yeah, yeah. Mostly mixing video. I used to lug my huge Macintosh quadra to parties and.

Oh, my God. Courageous.

Oh, God.

Oops. Yeah. Wow. Yeah. I also, in my first year of doing, like, multimedia things, I also wrote a VJ tool. I think at that time everybody had to do that.

Oh, yes, absolutely. If you were a Max MSP user, you had to write a VJ tool.

I did it in direct. So. Okay. Then you met Eric. How did you meet him?

I met him through, actually, my very first boss, Darren David, who currently has a company called Stimulant IO. And Darren introduced us when Eric, who was doing stamen solo at the time, was looking for somebody to help him with a project for BMW. So our first gig together was me as a contractor doing backend programming and data programming and xml for this kind of online decision making support tool for the design works group under BMW.

Nice.

So we got a few good trips out of that to go to Munich, got to see Oktoberfest for the first time, and then we hit it off and things worked great. And in 2003, I joined full time and he and I started growing and building and making good stuff.

Cool. Yeah. Interesting. So you never really, you never went to a university or have like a computer science degree or something?

I have sort of a computer science minor from UC Berkeley, you know, took all the lower division courses, did a bunch of human computer interaction stuff, such as it was at the time, and sort of new enough to be a little bit dangerous, but mostly focused on cognitive science.

Okay, cool. Yeah, interesting. And, yeah, 2003, you started with stamen, I think I first noticed stamen with the dig labs visualizations. Probably as many people, I guess.

Yes. That was a huge, massive opportunity for us. We had already actually been doing a little bit of mapping and fairly serious visual design work before that. Dig, I think, was the moment at which everything came together with online data streams and real time data. And the client, who was completely on board, was doing interesting work.

And, yeah, for those who don't know it, probably many don't know it because it's offline, but there was a series of visualizations around dig. We should explain dig. It's like, it was just. It was like Facebook, but just the liking. Yeah.

Like voting for news stories that would get pushed to the front.

Exactly. Yeah, yeah, yeah. So a bit like a Reddit precursor, maybe even, I don't know, between social bookmarking and just liking stuff. And you did a series of real time visualizations, or some real time, some, I think, longer timeframes on what is being dug at the moment. And they were very organic, very fluid, and just new, you know, nobody knew you could do this type of real time thing on social media.

Yeah. We had heard a lot of feedback at the time from journalists who said that they were basically setting up the dig swarm or dig stack visualizations on their second monitors and just kind of peeking over every once in a while to see what was interesting to write about and think about.

Perfect. Yeah.

So, Mike, I wanted to ask you, how did you start with the idea of stamen? Was it more that you started with a big vision or just with a single project and then this turned into a big thing?

Speaking for myself, it was a single project that later turned into a big thing. I think, for Eric, he came from a background working with a company called Quokka Sports. He was a creative director there, and they basically did kind of sports tech, innovative visualizations showing mountain climbs and motorcycle races and things like that. So I think he came from a much stronger kind of explicit narrative, creative direction background. For myself, the data was always the really interesting part that I was involved in. For me, the opportunity was to do something with live data online and something that pushed data streams online in a way that I hadn't really seen many people do before.

Okay, sure, sure.

Yeah. And for quite, quite some years, Stamen was basically the only data visualization studio. I mean, we discussed it last time because we had Santiago on the show, but there was. For a couple of years, there was only stamen at the west coast. There was Bestiario in Madrid, and that's it. More. And there were a few people doing it, sort of, like, independently, but, yeah, that's it.

Yeah. It was definitely kind of uneasy not having competitors because we were wondering if we were doing the right thing or not.

Like, awkwardly quiet around you. You're like, are we on the right track here?

Or did we stumble into an academic medium by accident?

Yeah, maybe. Yeah. Yeah. But then, I mean, stamen became, of course, mostly known for the mapping activities. Right? So it's been a occasionally little excursion into information visualization or, like, the occasional network or something like this. But I guess 80% of the, at least the well known projects were maps, and always very innovative maps. So how did that come? Was that also more like a sequence of events or was that a plan from the beginning?

I think it was a sequence of events, but it was something that we were doing very early on, before I actually met Eric, he had done a couple of map projects with MoveOn.org comma, which was a political action committee here in the United States fighting for left causes. He did a few maps that were showing kind of where activity was happening around the US, around marches against the Iraq war, things like that. And then in 2004, they asked us to work on a project that was sort of like an online call in show with a live map component that would allow people to see where in the country other people were at that point. We had just recently seen a project from IBEam in New York called Fundraise.org comma, which was not quite live, but up to the month live visualization of election finance information, basically individual submissions from people of money to. To various political causes. And it was the first time I had ever seen data mapped on a national scale in quite that way. So we did this project that was basically like a live call in show that used a real audio stream to broadcast audio, and then people would log in with their zip code to a flash map, be placed along with everybody else that was on the map at that same time, and be able to see if they were in a kind of crowded, left leaning county, Orlando, completely desolate, you know, nowheresville, up in Wyoming or something. But we actually designed a full backend system to translate that information into maps for everyone. And I think at its height, it had something like 50 to 100,000 people on this thing. At the same time, it was just a super fun project to work on.

Wow. Yeah. Cool. Yeah. And people. Yeah. It was so exciting. Or it's still exciting that people use maps these ways and create their own maps that are populated with stuff and that are constantly changing, you know, I mean, of course, also the things like Oakland, crime spotting come to mind, where I guess, also, for the first time, really, on that scale, police report data was visualized more or less real time, let's say, on a daily basis or so. I think that completely changed how people thought about maps, you know?

Absolutely.

So before, a map was something that wouldn't change, you know, you buy it, like, on paper and becomes maybe a bit dirty, but that's it. And suddenly maps become these lenses where you can look into the world.

Yeah. It's an incredible shift in how people interface with geographic information. I mean, especially now that you have things like Twitter, which for certain types of tweets will actually encode the latitude and longitude where somebody is. So you end up with a situation where you can actually live and in real time, watch people talking to each other in space. It's kind of an incredible shift from a time where you would have to collate and collect data over weeks or months and then produce sort of a flat or static thing.

Yeah. Do you think, can we reconstruct a short timeline of what has happened? I think it could be really interesting to give that an attempt. So let's say you lived before 2000.

I mean, nobody's that old.

People did that. People did that. I read it a on the Internet. Basically you just had paper maps, right? Or am I wrong? Maybe there were a few CD roms.

Yeah, a few CD roms. I mean, you know, absolutely. The 800 pound gorilla in the mapping space before this time was Esri, a company here in California. And what they did was they produced hardcore GIS software. You know, things like view sheds and hydro analysis and bathymetry and political boundaries, you know, all kinds of really, really high end, powerful, expensive tools for doing geography. What we started to see in the early two thousands, that was really fascinating to me, and this is even true from a job that I had before stamen, where I was working for a public relations agency, was that you started to have availability of open datasets. At the time, it was pretty primitive. It would be like, I don't know, a CSV file of cities and latitudes and longitudes around the world that would allow you to do a city scale map showing, you know, offices for a business or something. Pretty primitive stuff, but very open source and very open to the possibility of kind of hand tooling visualizations of that. When we did the move on work, I think I was actually doing my own sort of equivalent of shapefiles in Adobe illustrator in order to get a particular type of cartographic generalization, partially because the tools didn't exist, and partially because I didn't know about the tools that did exist in the burgeoning open source world. And then I think in 2004, everything just completely exploded. I mean, you had Michael Fruman and Jonah Peretti doing fundraise at I beam. You had the mapping hacks book from O'Reilly written by Skylar Joe and Rich. And then later that year, or maybe even the beginning of 2005, Google Maps came out and just completely changed the landscape.

And it's unbelievable, you know, it's only like seven or eight years. Can you imagine that? And before, you never even had a tiled map, you know, that where you would sort of just type in an address, it didn't exist. I mean, it's so crazy if you think about it.

Yeah, it was incredible. I mean, I was at a really interesting session run by a couple of the folks from Google at last year's North American Cartographic Information Society conference. And, you know, I think the cartography world has sort of a love hate relationship with Google, because what Google does is, in some ways, they kind of ran roughshod over a lot of accepted cartographic principles in the early years, but at the same time, they went from nothing to an entire country, and then later the world in the span of just a few months by essentially throwing away a lot of common assumptions of what the right way is to do. Maps.

Can you mention one?

I think the biggest thing is the spherical mercator projection.

I mean, it takes some guts to bring out a map that has this projection, but it's. But that's exactly the spirit. It's a move that is extremely efficient computationally, you know?

Exactly. Yeah. If you model the whole world as a big ass square, suddenly you can treat it like, you know, squares and images and files on servers, and it becomes just, like, so much easier to handle from an engineering point of view.

Exactly. Yeah. If you model the whole world as a big ass square, suddenly you can treat it like, you know, squares and images and files on servers, and it becomes just, like, so much easier to handle from an engineering point of view.

Yeah. And no cartographer would ever dare to do that. They would lose their job.

That was an energy. Right.

Can do it. And exactly. This hacking spirit, I mean, propelled the whole thing, probably.

So. Mike, do you.

But, I mean, now we are stuck with people who think that new fundlet is huge and Africa is a really small country. That's sort of the downside. But what can you do?

Yeah, it works when you zoom in, maybe less so when you zoom out.

Exactly. Yeah. And, I mean, that really changed the whole thing. And, I mean, in the beginning, how would you work with. With maps? Like you said, in the beginning, it's very much handmade, like illustrator and flash. Did you then later move on to customizing Google Maps soon or around that time, how would you produce your maps?

Usually around 2006 or so is when I first started to become aware of the OpenStreetMap project. So prior to then, yes. You would essentially just use Google Maps for a project at Stamen. We actually weren't doing too much mapping during that timeframe. 2005, six and seven is when we were doing the most of our kind of dig style work. So a lot of our projects were focused on time streams and social user streams of information, rather than cartographic stuff.

Okay.

The Google Maps aesthetic was one that I think made it really easy to do a lot of stuff on a street scale map of the world. But at the same time, it always had this kind of kooky sort of acid house look that didn't necessarily fit with what we were wanting to do.

Yeah, I could see that.

But around 2006 is when I got interested in OpenStreetMap. It's around the time when I was first exposed to the Mapnik renderer from the guys at everyblock. And at that point the data started to get good enough to where in 2007 we were able to do a Mapnik based visualization map for the London Olympic Games. This was very early in their planning stages, four or five years ahead of the games.

Yeah, yeah, yeah, yeah.

So that was, I mean, that was a really fun project, I think. You know, I was probably one of the few designers I knew who looked at that logo when they first unveiled it and thought, this is the greatest thing I've ever seen, and just totally bit into it and just loved the, like, magentas and greens and crazy colors.

Yeah, that must come from your past.

Yeah, yeah, yeah, exactly.

Actually, I agree. I think it's a corporate identity that worked in the end, but I didn't realize that until I saw it, like, while the games were on and so on. But, yeah, you must have been the only one. Pretty much.

That's probably true. It definitely blew my mind. I had never seen anything quite like it, and I just felt like there was a sort of, you know, I don't know, eighties revival coming that was gonna make it work.

Yeah. And so 2006, roughly, is when OpenStreetMap, did it take off at that time or was it around already earlier?

It was around a little bit earlier. Steve coast and Tom Carden and a few other folks had started it as a UK focused project earlier. I forget the exact year it might have been. 2005 or so.

Okay, got it.

Maybe even as early as 2004. But it was literally just looking at GPS traces from a courier company that had donated some data, trying to build something from London outwards. And then for us in the US, it became a viable source of data when in 2007, the entire tiger line dataset from the census was imported and suddenly it was something that you could make reasonable maps out of at a medium scale and just totally opened up the world for us. Yeah.

That brought with it also new set of tools and toolkits and services and so on, right?

Oh, yeah. No, I mean, this has been for my role, which was mainly kind of an R and D technology, data focused role. The most exciting thing about all this stuff was the fact that suddenly you had this entire community of people around the world that were working on tools for data and tools for moving data in a purely open source way. It was just such a radical shift.

Yeah, it's crazy. And also just the fact that all this mapping data is available suddenly and also open street map, many don't know that, but it has a lot of additional layers of what you don't. It's not just a street map, you know, in the literal sense, but you also have like all kinds of hotspots, bicycle paths, pedestrian paths, all kinds of overlays. It depends a lot on the country you're in or what the local community thinks is fun to do, but it's a huge resource, and I think I see it by now as important as Wikipedia, almost just from another angle.

Yeah, it's kind of a sleeper. We were talking about this Atlantic Cities article a little bit before the podcast where they're looking at some of the new projects that could inform how city data has moved around. And of the six projects that use maps on this page, five of them use OpenStreetMap. I don't get the sense that it's a very painful or deliberate choice. It's just the obviously good thing to use.

Yeah, it's kind of a sleeper. We were talking about this Atlantic Cities article a little bit before the podcast where they're looking at some of the new projects that could inform how city data has moved around. And of the six projects that use maps on this page, five of them use OpenStreetMap. I don't get the sense that it's a very painful or deliberate choice. It's just the obviously good thing to use.

Yeah, and it has become much easier. So in the beginning you have to be really Mike Migurski to do anything with it. But by now also mere mortals can do something fairly quickly.

Yeah, I mean, for me, I think I had to learn from Paul Smith, who was working at everyblock at the time. And, you know, that was the first person who ever sort of stepped me through what it meant to compile a really weird, cranky piece of software like Mapnik. I've never done anything like that before.

Can you say two words about sentences, rather about every blog? I think it's a really interesting project, and I heard it's sort of come to an end or has been cancelled.

Yeah, it's sadly come to an end. I mean, it's been influential for us at stamen in a number of ways. When we first launched Oakland crime spotting back in 2007, one of the projects that I was explicitly looking at for inspiration was Chicago crime, which was a project of Adrian Holavati, who then went on to found everyblock. All the people who founded everyblock, Dan, Adrian Wilson, Paul and others were hugely influential in visual design, graphic design, and information visualization on static web pages in a way that I think is still playing out today. So it really made me sad when MSNBC decided to kill that project because I think that they were doing just fantastic stuff.

What was like local news project or like a way to connect neighborhoods? What was the main purpose.

Yeah, it sort of transmogrified, I guess, into something that was more about connecting neighbors to neighbors, but when it first started up, it was about government data. So Dan O'Neill, who currently runs a project called Smart Chicago, was essentially their people person. And I, as far as I could tell, he spent the first few months of everyblock basically just moving from city to city to city, trying to convince city agencies to open up datasets. Whether it was like, I don't know, firetruck calls for service or restaurant health inspections or filming locations or sewer data or whatever, he was basically just knocking on every door in every city hall that he could find, trying to get people to open up datasets to every block. Really an early pioneer in that kind of stuff.

I think it's funny because now it's kind of like the contrary. Right. So before you had to convince people to make their data open, and now it's more kind of like the opposite. If your data is not open, you are basically behind the others. Right?

Exactly, yeah. I mean, you're seeing things like, you know, in Dan's home city of Chicago, for example, the, you know, John Tolva, the CTO, and Brett Goldstein, the CIO, have just released tons of city data using GitHub, which is fantastic. And directly from the source.

Yeah, that's also totally changed. Right. I mean, for Auckland crime spotting, you were still like, sort of competing with the police, maybe, or the city or the relation wasn't that clear, I think, for everybody. And that's just a few years ago, right?

Yeah, we had to do some pretty crazy scraping in the early days.

So can you tell us more about how the project started?

The crime project? Yeah, yeah, sure. So, you know, Oakland has its own website for disseminating crime information. It's called Crimewatch, and it uses a fairly primitive arc web based tool from a couple of years ago to display the data. But unfortunately, what the tool does is it forces you to step through a number of questions in order to see data. So you come to the site, you have to accept a user agreement of some sort. You have to tell it where you want to look, whether it's a zip code or around a school, tell it what kind of crime you want to look at, basically this kind of like, short litany of questions, and then it will show you a very primitive looking map of your data. Okay. We were thinking about that and thinking about how much more useful it was to look at things like Google Maps, for example, where you would have data just shown to you first before you were necessarily ready to formulate your question.

Moritz WhatsApp on your telephone?

I can mute it. I had to unplug it. It had to be done. Yeah.

But, yeah. Anyway, so we were looking at this site, crime watch, and basically thinking that there was a better way to interface with this stuff, a way where you could see the data first and then drill down rather than have to come up with a list of questions first and then see your data. And so the feedback we got from it after we launched was just wonderful. I mean, we had, I remember in particular, one guy mailed us from one of the downtown districts and said that their monthly meetings with the police had totally shifted where previously the police would tell them what had happened, and they would have to kind of process and respond to that at the time. And then subsequent to crime spotting, they were able to come to the police and know exactly how many robberies or car thefts there had been. And so the conversation could be sort of more productive on a databases just right off the first step.

Wow, that's huge. So that's a real success story of visualization.

Exactly.

I don't know many other success stories like this. So it's.

No, it's really one of these things that, where you can also say the visualization actually changed how the whole city works, I think, you know, in a way. And that's really, it's quite powerful to open up all this data.

Yeah, that's a more.

Yeah, it's interesting to me. Sorry, go ahead. No, no, go ahead.

Yeah, it's interesting to me. Sorry, go ahead. No, no, go ahead.

Yeah, I mean, that's a general, I think, interesting problem or issue in visualization that there are so many cool projects around, but it's very hard. I think a very tiny proportion of this project can actually tell such a nice story like the one that you just said. Right. I mean, having a real impact on what people do or decision that they make or improving their life or whatever. I mean, it's very hard to go from a cool project and boiling it down to having a real effect on the world. Right.

Yeah. I think what's interesting about that story, too, is that it's a story that doesn't lose meaning when it's repeated from place to place. You know, every, every single city in the world could benefit from a story like that because every single police district and zip code and, you know, census tract in the world has people in it that care about their neighborhood and want to learn about their neighborhood.

Yeah, yeah, yeah. I think that's a general characteristic of some visualization, when some visualizations, when people can actually relate personally to the data that is shown, then it's. It has a completely different shift.

Right, yeah, exactly.

And are you aware of. So I cannot recall. So crime spotting was only for Auckland. Right. And you never extended it to other cities or.

We did actually extend it to San Francisco.

Okay.

And there was a fairly interesting process there where, with Oakland, you know, we got interested on our own. I spent an entire Christmas break basically trying to beat the data out of this crime watch site and then worked for a few months afterwards kind of collecting and collating and figuring out how to map it, and ultimately released it as a studio project in summer of 2007. Later on, the city of San Francisco actually approached us. They were putting out a data repository called datasf.org dot, and we got contacted directly by the city to release a San Francisco version of crime spotting.

Yeah, but that's exactly the change. Right. You have to demonstrate once you know what that means and that it's good and then people start to get it, but you have to show it once.

Yeah, exactly. It's frightening.

Against the odds, sort of. Yeah.

Yeah. The process in Oakland was sort of a funny one because we had to kind of almost go spelunking for our own data for a few months, and then the city shut us down for a little while while it tried to figure out how to respond to this. And then ultimately the city came to us and said, okay, fine, this is great. We like it. So let's figure out how to get you guys data more reliably.

They didn't arrest you?

Yeah, totally.

Listen, we put here in jail for three weeks, but now we thought about it and actually it's kind of cool.

We love the thing we made.

Exactly. Can you make some more crazy? Shall we talk a bit about tools and frameworks and so on? I think that's a really interesting area too.

Exactly. Can you make some more crazy? Shall we talk a bit about tools and frameworks and so on? I think that's a really interesting area too.

Yeah, sure.

I mean, you put out modest maps also around that time, which was, I think, really helpful to many people because it made this, like making your custom map quite painless. So basically you could use it in flash first, I think that's right.

Yeah. It was an actionscript two and then actionscript three at first, right.

You could load like Google Maps or Microsoft Maps or Yahoo Maps and have little map controls and overlay your own markers and so on. So I did a few projects with that myself and it worked really well. Is it still a running project, modest maps?

Yeah, it is. We were recently contacted by the Flex project, which is now under the Apache foundation, and agreed to donate the actionscript code to Apache.

Nice.

The JavaScript code is still in heavy use, especially by the folks at Mapbox. They've been building a lot of their browser based interactive tools around a combination of modest maps and leaflet, which is another JavaScript library.

Yeah, yeah. There's also unfolding by Till Nagel, and I know he started with modest maps, but then sort of transformed it so much that it's not really modest maps anymore, and that's a processing version, let's say, of the same idea that you can use map tiles from other providers and do your own stuff with it.

Yeah, exactly. I think it was in doing that project that I realized that all the three major providers at the time used the same map tiles and the same.

Logic, more or less just different naming conventions.

Yeah, exactly. I thought it was all going to be different and realized that basically for every single one of them the world was the same square.

It's interesting then I guess polymaps came out from Staman as well.

That's right, yeah. So Mike Bostock, who is better, more recently known as the creator of D3, worked with us for a summer.

Oh, that's crazy.

Yeah, you can see the signature in his code whenever you see that kind of the method chaining stuff, you know, that he's had a hand in it. And that was basically, you know, Mike has an interesting point of view about this stuff, because his basic view is that browsers and JavaScript and SVG are the undisputed future, so he might as well be living in the future right now. Yeah, and so polymaps was really kind of a stab at the future. It was an attempt to basically say, okay, if we can assume fast JavaScript and high quality SVG rendering, what can we do with maps? Turns out what you can do is very, very good in browser vector rendering, which no one had really thought too deeply about until that point.

So how does it work? Do you actually load the bigger shapefiles into the browser, or do you have portions? Is it like vector tiles? Is it like that?

It could do a combination of them. So you could either just load one big Geojson file, which was a fairly common use case, or in the demos that would do things like let you zoom all the way down to county levels. We had vector tiles, so tiled Geojson, that allowed you to publish different resolutions of data and different generalizations to that on the fly.

And you could also use bitmap tiles, right?

Yes, exactly. So that was one of the first frameworks that let you kind of put those two things together. So you could have, like, maybe an open street map based background, and then I. Clickable blobs that would change shape or do something interesting in the foreground.

Exactly, yeah. And I think this layering came there for the first time, that you don't possibly just have one base map and then some markers on top, but maybe like a heat map level in between and a few more shapes so that you build really complex maps on the fly. More or less.

Absolutely.

Yeah. Now you have this whole ecosystem around OpenStreetMap, basically. So there was cloud made pretty early, I think, as well.

Yeah, I think, I want to say 2007 or 2008, maybe.

Yeah. Which allowed you to customize map tiles.

Yeah. And we also did a number of their built in default map styles, kind of based on our experience of putting data and maps together. So we created one called Midnight Commander and one called Pale dawn that were basically the sort of dark and light versions of the background map that you wanted to have your data in front of. And interestingly enough, we still see people using those on a fairly regular basis.

Yeah. And I think Tilemill also has a default that is very close to Midnight commander.

Yeah, that's right. Yeah. Like the really dark version. I think they've been talking about the zombie map they did recently. That's. It's almost like the red Death version of Midnight Commander.

That's nice, too.

Yeah. The creative brief from cloud made for Midnight Commander was what kind of map would Jason Bourne use on his in car gps?

But that's the nice thing that you can really start to play with maps in that way and give them a certain styling, a certain look. And also cartograph, which came out one or two years ago by Gregor Eisley, which was also fantastic because you had these demos that looked like really vintage maps, but at the same time you realized, oh, a part of it is animated. It's fairly dynamic. I can see that. So that's before you just had standard maps, and now it's the whole design space opening up, right?

Totally, yeah. I mean, when you have that level of data available to you, you can do a lot of really kooky stuff with it.

So if you have a new, let's say, map project, let's say you had Oakland crime spotting, you would start that today. What would you use? What toolset or what combination of tools?

Good question. I guess it would depend on what I was trying to do with it. If it was something like Oakland crime, probably the first thing I would reach for would be leaflet j's. I think it's probably the best supported, easiest to use, out of the box mapping solution framework library thing out there currently, yeah. If I wanted to get weirder, I might reach for polymaps. I've been working on a framework of my own called squares recently that addresses some of my own thoughts about how to get Webgl involved, but that's still pretty early, so I don't know if I would necessarily reach for something like squares then. Modest maps is just the kind of hyper minimalist approach which basically just throws tiles on the screen and forces you to do all the other parts yourself.

And in leaflets, can you also have different layers of data overlays, or is that more like tiles and markers, really?

You can definitely do different layers and overlays, yeah. You can change opacity. You can lay multiple bitmap tiles on top of one another and have them show through. I did a project recently called green means go around us census data coverage in the US with OpenStreetMap and used leaflet. It was shockingly easy to make something that looked exactly how I wanted it to.

Yeah, that's nice. And let's say if you wanted to process lots of data, like, I don't know, county level information in the US, or even more like, finer information, and do a lot of, maybe render a really high res map, how would you approach that?

It sort of depends, I think, if you wanted to render a really high res map. I think probably what I typically do is custom Mapnik styles. I tend to skip tile mill and go straight for cardo, CSS or XML under the hood because it allows me a little bit more flexibility in how I've always worked on the stuff. But one of the things that I'm really excited about with Mapnik recently is they've started to support a scale factor applied to the entire map. So you can start to do things like designed for the screen, double resolution for retina tiles, and then start pumping up the resolution even higher for print tiles.

But you use the same assets and the same base description, but you just change the DPI more or less.

Yeah, exactly, exactly. So it bumps up all the font sizes, makes all the lines thicker, makes it all print style.

Oh, nice.

Yeah.

And mapping is like a server software, or you can run it on your machine, probably, and it will take care of the rendered rendering of the map.

Yeah, that's right. Yeah. It was built for OpenStreetMap. Artem Pavlenko from Cambridge in the UK first put it together in about, I want to say, 2006 or so, possibly even earlier and it was one of the first kind of server map libraries that used extremely high quality anti aliased rendering for fonts and lines. And the output just looked miles better than anything else that was available in the open source scene at the time.

Hmm. And, and I know only matte box and tilemail and I found that really impressive already that you have like a CSS for maps and all these different like conditions you can include like on zoom level five. My font should be italic except the ones for cities that you know. So that's really quite, quite impressive. But is that more or less just let's say just a wrapper for the more complex, complex Mapnik functionality? Could you say that or did they add something generally new?

No, it's definitely a wrapper. I did a project about, I don't know, four years ago called Cascade Nic, which was a kind of precursor to cardo CSS, where I had been working with Mapnik XML style previous to that. And once you're doing something fairly complicated, it's really easy to get in over your head with, with XML stuff. So I designed a CSS syntax which compiles down to Mapnik. It basically turns itself into XML. But the idea is that you can express things in a way that's declarative, the way that you do in a web browser for HTML, essentially. So tilemill and Carto CSS borrowed that idea and built on top of it to create something that was much, much more accessible than the old XML to a much larger audience.

Yeah, that's the beauty of it. You can just open an example file and change a few parameters and boom, your map looks different. I think that draws a lot of people into the more complex stuff then.

Definitely.

Have you played a bit with D3 and the new geo plugins and the 500 map projections thing?

Oh my gosh, yes, I played with it. Yeah. Actually the library I was talking about earlier that I've been working on called squares, uses D3 under the hood to do all of its tile placement and basic dom mathematic. D3 is amazing. I think it's been really interesting to see how people react to it because it's such a virtuoso piece of software. And I have the sense that it's taken people two or three years of watching it evolve before the kind of core power of the thing, the join syntax in the middle, is becoming understandable to folks.

Yeah, it's a whole new paradigm suddenly.

Absolutely.

And always when I, when I try to explain it to people, sometimes when I'm in a good mood. I seem to get it myself, but sometimes I explain it and I just realized it's a bit unexplainable. You just do it like that and then you'll see what happens.

Totally. Yeah. I went to a meetup in Denver a couple of months ago of JavaScript Geo folks, and D3 was the star of the show. But I kind of walked away with the sense that almost nobody in the room had ever really understood it. So there was a tutorial where the person running the tutorial was like, well, I just figured this out myself, so let's see if I can explain it to you.

Probably that's the whole magic of it.

But it's amazing. I mean, it's incredible the reaction that people had with D3. I mean, I think it's basically becoming the standard right right now, totally.

But you know, it's the 8th most sard project on GitHub also, you know, it's in front of Ruby. I'm not kidding. Not Ruby, but Ruby on rails.

Oh my God.

And that's so crazy. But I think at the same time I think the interest is huge, but I'm not sure if the actual use is at that level yet.

Oh, that's a good point.

Yeah.

Yeah. So I think everybody's like super excited about it, but I think, let's say professional applications or services where it is under heavy use already, it's probably limited, but I mean, it's totally gonna happen. I mean, there's no doubt about that. But I think at the moment everybody is just psyched what you can do with it because Mike is demonstrating so nicely also what you can do with it.

Yeah, well, I mean, both Mike and Jason Davies have an amazing approach to it where they're just, I mean, especially with the geo stuff, they're figuring out a lot of issues that I think a lot of developers are frankly never going to think about. Polygon clipping and line simplification and these kind of very core issues of graphics programming that are what make D3 work as well as it does.

And they just solved that and put out a demo.

Yeah, put out a block. Super easy.

Yeah, that's fantastic. And where do you think, where are things going like? Or what are the big challenges at the moment? Where are the gaps? What doesn't work, what should work?

Yeah, that's fantastic. And where do you think, where are things going like? Or what are the big challenges at the moment? Where are the gaps? What doesn't work, what should work?

I think the biggest gap, the biggest opportunity, and the most exciting thing that I'm seeing right now is the new availability of data and the increasing availability of data in forms that are easy for people to digest and consume. So one of my research projects over the past few months that I've been sort of unemployed and experimenting with stuff, has been thinking about how a dataset like OpenStreetMap can be available to people who are not database jockeys on their own. You know, currently, for example, like if you use tilemill and you want to style a base map, there's a huge slice of stuff you need to know about under the hood in order to get to that state. You have to understand post gis, you have to download the planet file. You know, it's almost 30 gigs today. It's just colossal stuff and a gigantic pain in the ass. So one of the things that I've been exploring for the last couple of years is how can you get data that's open and produced by this community into a form where it can be sort of nibbled on in small chunks and approached in small ways by people who want to make something local? So a few years ago, I started out something called the Metro Extracts project. Wherever I chose cities around the world and created extracted versions of OpenStreetMap shapefiles just for those cities, so that if you lived in Chicago or New York, you didn't have to go dredging data from the open street map server. You could basically just download something really quick and use it.

Is that the data sets up on Cloudmade now?

Cloudmade, yeah, they do extracts in OpenStreetMap form and a couple of other forms. The metro extracts are slightly different. They're ones that are more about metropolitan areas. For example, a Kansas city that spans over two states will be in there.

But there's no real tool at the moment where, let's say a mouse click person could say, I just want northern Germany and this type of information and give me that slice of data to download.

Right, exactly. So the other thing that I've been working on over the past couple of months has been something that I'm calling Mapnik Vector tiles, where I'm thinking about how do you make the data of OpenstreetMap available as a tile? Something in a kind of geojson like format where Mapnet can use it directly. And I've actually had some pretty good success with it. I'm publishing layers of streets and street labels and buildings and green spaces and actually generating good looking maps out of this stuff.

You're also generating funky Webgl maps I've seen on your blog that have rainbow colors and all the good stuff.

Is.

A really interesting perspective, of course, for rendering.

Oh my gosh, yeah.

What's your progress there. Are you, like, is it going to work soon, or do you think it's going to take a while?

Well, if you're into Mario Kart style rainbow maps, I think it works for you today. But, yeah. More seriously, I think, like, with a lot of labeling and cartography projects, the real hard part is in the labeling is in the text labeling, and that's something that I haven't really thought about in Webgl yet. I'm still just thinking about how to get the data into the browser. So it's probably still a few months off. I'm sort of hoping that somebody else starts to pick it up and pushes it forward a little bit, that you.

Meet some labeling expert on a conference.

Exactly, yeah. Somebody who's got a better grasp of geometric algebra than I do.

Talking about challenges, I mean, that's still a challenge. Right? Like, good automatic labeling of, like, any map constellation you want, you know, you want to display.

Yeah, I've been working on that for a few years with Nathaniel Kelso, who's a colleague of mine from stamen. We have a project called Dymo together that's sort of a labeling engine that focuses on cities and countries at the medium scale. Uses some pretty cool math that I found called simulated annealing in order to place the labels in a kind of more organic way.

Yeah, that's like machine learning, really. Like optimization algorithm, right?

Yeah, exactly. Yeah. Lots of fun. It's pretty dorky. It's hard, I think, for people who haven't really run into the problem. It's hard to imagine why. It's an interesting problem to work on, but I think with cartography and Dataviz, a lot of things are that way. You really have to kind of dig deep and go off into the weeds someplace and learn something really arcane before you can bring it back and make it interesting to folks.

Mm hmm. Yeah. And you have to think a lot in systems and tools. You know, you cannot just solve the problem for once, but you have to solve it on the general level.

Yeah.

And the earth has a lot of different places look a bit different, too.

Yeah. It's a big tension in the work, you know, how do you solve it individually for a single map, or do you solve it generally for the whole world? And, you know, do you end up off in some, you know, off on some vision quest somewhere, solving it for the whole world where it's not useful?

Enrico, you have more questions?

Let me see.

We can move to the tweets, otherwise.

We can move to the tweets.

Yeah, we had three tweets. So first one is from Scott, aka Aland, left. He asks, I don't know how, seriously, but he asks, what is a map tile?

Mike, why is the sky blue?

Exactly. Fruit flies like a banana.

Well, a map tile is so if you can imagine the world as a giant square image going from all the way almost at the North Pole to all the way almost at the south pole, and from east to west, then you can cut that image into four smaller images that are also squares and continue all the way down to street level. And what a map tile is, is essentially a standard for how you cut the world up so that you can layer images on top of one another and get a nice looking map out the other end. It has some particular benefits for the engineering of this stuff, because it means that when you render a particular map tile, let's say, for part of a town for one person, then that same image can be cached and represented to somebody else later on. So it's essentially the way that Google and the company they bought to build Google Maps, and then everyone since then has made it possible to do worldwide map coverage by essentially repeating these tiny little square PNG and JPEG images for people.

That's a great explanation.

I like it.

And it really, I mean, this whole concept, it's really admirable. I think, you know, we take it for granted now, but it's. It's really smart the way they set this up.

Yeah. Map tiles are the MP3 files of geography.

Yeah. In a way, yes. I mean, that's true. Sometimes a smart hack propels, like, a whole industry. Yeah. It's crazy.

So do you know who came up with this idea first?

You know, I don't exactly. The earliest reference that I've seen to the idea of a tile in an image is from a paper on a technique called MiP mapping from 1983, I think. I don't recall the name of the author, but I'm almost totally certain that he works for Google now. And essentially, it was a technique for very quickly rendering and pre scaling imagery so that a graphics card could represent imagery at different scales.

What's the name of the technique again?

Mipmap. M I p. M a p. I'm gonna see if I can find the author.

Yeah.

Lance Williams.

Lance Williams.

Cool.

We should definitely research that, because if he's actually the guy inventing that, we should maybe send him a bottle of whiskey. We kind of owe him.

Yeah.

Thanks, man. Thanks, Lance. Nice idea.

He's responsible for a major shift, actually.

Yeah.

I think a lot of interesting techniques in computer science are like this, where, you know, if you can figure out a smart way to trade processing for storage, then you can reap huge benefits from it. And that's essentially what map tells do. They basically say, instead of doing things in the old Mapquest pre tiles way, where you're rendering a special map for every single person who walks in, you can render a single map that's shown to everybody together, and then as a result, you can do much more complicated things with that map anti alias text and make the lines beautiful and do some crazy watercolor stuff like we did at stamen, or a kind of simulated hand drawn pen thing like CartoDB has tried. The techniques are pretty astonishing.

And you can sort of cache the computation results and represent them. That's really smart.

Storage. Cheap.

Yeah. Getting cheaper. I heard next week would be Jan Willem Tulp. He asks about process and where you get your inspiration. Like, how do you come up with ideas? How do you realize them?

Yeah. Getting cheaper. I heard next week would be Jan Willem Tulp. He asks about process and where you get your inspiration. Like, how do you come up with ideas? How do you realize them?

Oh, interesting. Yeah. These days, I think most of my work these days is focused on how data moves from one place to another. And so a lot of my inspiration comes from perceiving situations where there is an obvious consumer and producer of data that aren't talking to one another, or it's difficult to talk to them. I mean, the example that we were just talking about was things like Mapnik vector tiles for open street map data, where you have all these cartographers running around with GPS units collecting data out on the streets, and then people who want to do something with that data and present it in some way. And the mountain between them is this colossal planet OSM file sitting in the middle that needs to be chewed up and made smaller and made easier to deal with.

Yeah. And, you know, both sides. And so you can see in your head, probably, you have the match already developing, and then you're in the middle of another self initiated project.

I hope so. Yeah. I mean, so much Dataviz is like that, too. It's just, you know, it's the realization that there's something about some chunk of data that's going to be interesting to somebody out in the world who doesn't know that it exists.

Exactly. Yeah. Yeah. Some things just need to be done.

Yeah.

And third question from Jan. I mean, he really, he put three questions. What tweet? How did that work? What do you think is the trend in online maps?

I think we're going back to vectors and custom rendering stuff. If the new projection work from Mike and Jason and D3 is any guide or the new experiments that I and others are doing with vector tiles for datasets like OpenStreetMap or any guide.

The Apple maps are not famous for their quality, but they also use that technique. And, I mean, principal. It's smart, right?

Yeah.

Also on mobile phones and all these different devices, to have something that's sort of screen independent in a way.

Yeah. I mean, I think. I think the big change with mobile phones is that now everybody's walking around with a very high quality gpu in their pocket.

Yeah.

And so you can start to do a lot of that rendering on the client side and on the device in a way that you just couldn't do two, three, four years ago.

Yeah, it's interesting. Third question, Peterson, GIS, what is your approximate ratio of time spent on data manipulation versus visualization?

Yeah, it's interesting. Third question, Peterson, GIS, what is your approximate ratio of time spent on data manipulation versus visualization?

It depends on if you think those are two different things.

Yeah, I was just wondering where to draw the line there.

Yeah, I don't know. I mean, I think at stamen, I would say definitely it was like nine to one. My position was all about making data easier for our designers to work with, which in some cases meant, you know, pushing data towards them. In other cases, it meant helping them pull data towards them. But in all cases, it was about kind of breaking down that barrier between interesting data here and the technique to show it there. So, yeah, I would say it's probably like 90% of my time is spent schlepping data from one place to another and changing it from one form to.

Another, massaging it, yelling at it, coaxing.

It, waving treats at it.

10% color palette picking, right?

Yeah.

Yeah, it is. I mean, if you really want to get under the hood, it does become quite technical, I can imagine, and also quite challenging with all these huge data sets and different formats and.

Yeah, it's definitely easy to get stuck. I think if there's one thing that I wish I did more of, is the visualization side of things, is the actual tweaking of colors and representation formats and things like that to do the real visualization and graphic design work. But my side definitely tends to focus more on how the data pushes that.

I think it's funny because every time I hear to this question, and I always get a very similar kind of ratio as an answer, and I'm wondering if it makes sense to think about visualization without. So I would, why do we talk about visualization as separated from data manipulation? Basically, you can't really separate them. Right.

Interesting. Yeah, that's pretty true.

Can you do data visualization without data manipulation. No. Right. I mean, in an abstract way, yes. But in practice, even the way you design a visualization is heavily affected by the way you manipulate your data and what kind of shape it has. Right. So I'm more and more convinced that there's no data visualization without deeply thinking about even the data shape that you have.

Sure. My students are always shocked when I show them Ben fries pipeline, and, you know, there's this represent step, you know, and I tell them, you think this is data?

Yeah, exactly.

There's eight more steps you have to learn. It's like acquire, filter, part mind, you know, represent, refine, interact, and all of that has to come together.

Yeah, that's the thing. We end up debating for ages about what's the best mapping, the best color, the best whatever. But that's a tiny, circles are okay, yeah, exactly. Circles are pie chart versus bar chart and stuff like that, which is of course important. But in the end, in practice, the data manipulation side of visualization is so important.

Yeah, I think I've become convinced over years that the word visualization is always going to refer to the 10% at the end. I don't even know if that division is ever going to change, because in my mind, visualization basically means the last 10%. 20 or 30 years ago, visualization meant laying data out into a spreadsheet form, and then it meant laying data out into a pie chart form. And today it might mean laying data out into like, I don't know, a timeline or a tree map or something else, and it'll mean something else in the future. But it always refers to that last step of making something presentable for the public versus what you do for yourself to help you understand what you're looking at.

Yeah, yeah, it's true. I mean, the other thing I really always liked about your work is that you think in these really, in this longer process flows, really, like also in the projects, like walking papers or field papers, where you actually think about, okay, how can people collect geographic information or map information on a lo fi level, you know, like how to process that, like on paper, and how do we process it on the computer? And so that's the other part, like really building concrete tools and then see what happens with that and how people use it. I think that's a really exciting perspective in this area too.

Yeah, those are an exciting few projects to work on. Sorry, I'm a visitor here.

That's great anyways, because I think we're basically through, you have to run in eight minutes, so that's a good timing anyways. And, I mean, we have super much input. We'll have a lot of work to collect all the links of things we talked about.

Absolutely. I'll send you guys all the links.

Yeah, we might skip a few, though. It's a whole encyclopedia of maps. But it was a fantastic conversation and a great overview, I think, especially for those getting started in the field who might not be aware of all these projects in the background.

I hope it's not gonna be overwhelming. We collected such a huge list of links.

Exactly.

I mean, it was ten years of the most exciting ten years of cartography.

What can you do?

Ok, thanks so much, Mike. Fantastic catching up with you. I'm also super excited to see what you're coming up with next and what you're working on.

Yeah, likewise. And thank you so much for the opportunity to be on the show.

Yeah, thanks for being with us.

All right, guys, until next time. Bye.