Peter Batty is a co-founder and CTO of the geospatial division at Ubisense. He has worked in the geospatial industry for 29 years, and has served as CTO for two leading companies in the industry (and two of the world’s top 200 software companies) — Intergraph and Smallworld (now part of GE Energy). He served on the board of OSGeo from 2011 to 2013, and chaired the FOSS4G 2011 conference in Denver. He serves on the advisory board of Aero Glass.
Peter was interviewed for GeoHipster by Mike Dolbow.
Q: How did you get into GIS and/or mapping?
A: Totally by accident. I studied maths (as we say in England) at Oxford, then stayed on to do a Masters in “Computation”. I took a job at IBM, which was the dominant computer company in the world back then, in 1986. It was a bit like going to work for Microsoft in the 1990s or Google today, but distinctly less “hipster” and with more blue suits. On my first day they told me that they were introducing this new product from the US into the UK, and gave me the manuals and told me to learn about it. It was a product called GFIS, for Geo-Facilities Information System, a GIS focused on utilities and telecommunications applications. By the end of the first week I was the UK expert as I was the only one who had read the manuals, and I’ve been in that field ever since. GFIS ran on IBM mainframes and used specialized high-end graphics terminals which cost around £25,000 ($40,000). We’ve come a long way since then!
I found it a really interesting mix of challenges from a software design and development perspective — lots of interesting database problems, a very graphical focus obviously, which I find appealing, the challenge of designing systems that can be easily configured and customized, and more.
One thing I might mention from IBM days is that I was one of the main advocates in the industry back then for the notion that you should store all aspects of your geospatial data in a single database management system. This was the approach that IBM took, but it was uncommon in the industry at the time. The other main company pushing that approach was a Canadian outfit called GeoVision, led by Doug Seaborn. Esri’s product at the time was Arc/Info, the predecessor to ArcGIS, and it was so called as “Arc” handled the graphical / geographic aspects of the data, while “Info” stored the alphanumeric data. Pretty much all systems back then handled the graphical and alphanumeric aspects of the data separately. I wrote a paper called Exploiting Relational Database Technology in GIS, in 1990. In 1995 by chance I went to the launch of what is now Oracle Spatial, but was originally called Oracle MM (multi-media), at a conference in Vancouver. I chatted with the development manager and he said that they had a copy of my paper posted on the noticeboard in their office, and it had been a big influence on them. Which I thought was cool (though I’m firmly a PostGIS person these days)!
Q: So what did you do after IBM?
A: From IBM I went to work at Smallworld, who were a fairly early stage company at the time — I think I was employee number 30 or so, and after a year with them in the UK I was the first person to move to the US when we started the company there, and I’ve been in Denver ever since then (that was 1993). I imagine many GeoHipster readers may not have heard about Smallworld, but we revolutionized the GIS industry in the 1990s with a lot of ground-breaking ideas, and became the market leader in GIS for utilities and telecoms — you can read more in a blog post I wrote on the occasion of Smallworld’s 20th birthday.
GE bought Smallworld in 2000 and as with many acquisitions, it was a big culture change, especially for for those of us on the management team. Four of us left to form a company called Ten Sails, which evolved into what is Ubisense today (more of that shortly). In 2005 I took a detour to spend a couple of years as CTO at Intergraph, who were the second largest GIS company at the time, with revenues of around $700m. I had a good experience there working with lots of great people, but in 2007 I decided I wanted to get back to being more hands on with technology and left to start my own venture called Spatial Networking, where I did a variety of interesting consulting projects and also built an app involving social networking and future location called whereyougonnabe. I haven’t looked at this in ages, but I just found that our original web site is still out there, and there are a couple more videos here. I still think it was a cool idea, and don’t think anyone has really implemented what we came up with. Dopplr was the closest thing, but they didn’t have a lot of the geospatial features that we did, and they were acquired by Nokia and subsequently disappeared. I occasionally think about revisiting this idea! But anyway, we didn’t get the traction we hoped for with it, and in 2010 I decided to rejoin Ubisense.
Q: So you went back to Ubisense as CTO of the Geospatial Division. What are some of the more interesting projects you’ve been working on lately?
A: Well one half of Ubisense does RTLS (Real Time Location System) applications, using our precision indoor location tracking technology to power various applications, mainly focused on the manufacturing space. It’s a very cool technology — and I’ve come to learn that indoor location tracking is a surprisingly hard problem to solve!
The other half of the company is the geospatial division, which is where I work. I led a skunkworks project in 2010 to build a product called myWorld, which has really taken off over the past few years. It’s a web and mobile geospatial platform focused on large utilities and telecom companies, the same space I worked in with previous companies. It’s built on open source components — PostGIS, GeoServer and Leaflet being the main ones. Our mantra is “Simple, Smart, Fast” — it’s targeted at the 95% of people in these large companies who aren’t GIS people but can get a lot of value from a really simple Google Maps style interface onto their enterprise data. We’ve had a great response from customers. In fact one of them, one of the largest cable companies in the US, does a satisfaction survey of their end users every time they roll out a new IT system, and myWorld got the highest rating of any application in the 20 years they’ve been doing this, so that was very gratifying! We’ve been doing a lot of work with offline systems for use in the field, syncing data from PostGIS to SpatiaLite — since you still can’t guarantee having a wireless data connection 100% of the time, our customers really need the ability to work without a network connection. We have the core offline capabilities rolled out in some large implementations, and now we’re working on some interesting ideas for hybrid online-offline systems that can really simplify deployment and administration compared to traditional offline systems. We’ve also been building a number of specific applications to address particular business processes, like damage assessment, inspection and maintenance, and outage analysis.
One thing that differentiates us from the traditional GIS vendors is our focus on simplicity — it’s deceptively hard to make useful enterprise applications that really are simple and intuitive for end users. And another is that we are GIS-agnostic — at one of our large utility customers, we integrate with back end systems from GE Smallworld, Intergraph and Esri, and provide a common front end to all of them. The big players tend to work well with their own systems but less well with their competitors.
So it’s been fun. I feel we’re well on our way to disrupting the enterprise geospatial market in large utilities and communications companies, which hasn’t really happened since we did that with Smallworld twenty years ago.
Q: You presented your Geospatial Revolution talk to Minnesota’s GIS/LIS Consortium Conference in 2009. This was an influential talk for me and many of my colleagues here. It was the first time I had heard of the concept of “neogeographers” – did you coin that term?
A: No, it’s not my term, it was quite a widely used term in the industry at the time. At the time I would have said that Andrew Turner coined it — he literally wrote the book on neogeography, which was published by O’Reilly in 2006. However, I asked him if he came up with the term, during this interesting panel discussion at the GITA conference in 2010, and he said no, it originated in 1912 as a term to contrast with paleogeography. He credits Di-Ann Eisnor, then with Platial but more recently with Waze, as coining the recent usage of the term, though Wikipedia attributes Randall Szott with being the first to use it in the contemporary sense.
Anyway, it definitely wasn’t my term, and to be honest it’s not a term I really like, it was just the most widely used label at the time to describe the newer generation of systems that were disrupting the geospatial industry.
Q: What compelled you at the time to compare “paleo” or “neo” geographers?
A: Before I answer that, let me just digress a bit more on the terminology. One of the reasons I don’t like the term neogeography is that I don’t regard myself, nor most of the people doing interesting geospatial applications these days, as any kind of geographer. I’ve never studied geography at all, I’m a software developer and it happens that most of the applications I’ve worked on involve some aspect of location data or maps, but they involve a lot of other things too. Paul Ramsey uses the term “spatial IT”, which I like much better, though it still overemphasizes the spatial part in a way.
An analogy I like to use is that we don’t use the term “numerical information system” just because an application contains numeric data. We don’t have conferences on NIS. It doesn’t make you a mathematician because you develop an application that uses numbers.
Increasingly now, geospatial data is just another datatype, a map is just another aspect of a user interface. This opens up both the development and usage of geospatial applications to a massively broader audience, which is the real significance of the “neogeography” change or whatever you want to call it, the label doesn’t matter. Traditional GIS is a tiny portion of the geospatial ecosystem these days, I really dislike the whole attitude that GIS is this specialized sacred thing, and that you need to be a trained “GIS Professional” to do anything with geospatial data. Nonsense!
Q: Did the word “hipster” ever enter into your mind when considering either camp of geographer? As a Brit living in America, do you have a different take on hipsters?
A: Um, no! Was hipster a thing in 2009? And like I said, it wasn’t my term, I was just using terms that were prevalent at the time.
I have no claim or aspiration to be a hipster. And I’m not nearly as old as Steven Feldman so don’t qualify as a geohippy I don’t think.
Am I a GeoHipster? I had to consult your poll to decide on that, and according to that the joint leading characteristic of a GeoHipster is that they never refer to themselves as one, so it seems as though the only possible answer to give is no, whether I am one or not. But on your other top characteristics I score pretty highly: geoJSON is often the answer, I fired up ogr2ogr just yesterday to convert a shapefile into a 21st century format (geoJSON, naturally), and I have written my own code to roll map tiles (from Smallworld). I don’t think PostGIS is too mainstream though, I think PostGIS is fantastic. We’ve built myWorld on it and it has worked exceptionally well in some very large and challenging enterprise applications. I’d like to think we’ve been doing our bit to make PostGIS more mainstream — we have managed to get it installed in a number of large utilities and telecom companies where Oracle is the dominant database platform. With most of the customers that I work with, PostGIS would be regarded as very GeoHipster, I think. So maybe I at least score half a point there.
Another one of your poll answers was: “You call him Jack and still hate his company”. Well yes, obviously! I say that rather tongue in cheek, I’m not really a hating person, and I have plenty of friends at Esri. But I have spent my whole career competing against Jack, Esri has always been the “dark side” for me and the companies I’ve worked at. I have been in the geospatial industry for 29 years and have never used ArcGIS , does that qualify me for any sort of special GeoHipster award? Though having said that, we do use the Esri Leaflet plugin in myWorld now to enable integration with ArcGIS Server and ArcGIS Online, and quite a few Esri customers are using myWorld. Good interoperability with Esri is a strong focus for us at the moment.
So anyway, you be the judge on the GeoHipster front. I think maybe I’ve talked myself into getting a GeoHipster T-shirt.
Q: You’ve long been big on usability testing, even pointing out with self-deprecating humor how user testing showed a fatal flaw in one of your initial user interfaces. Do you have any more stories about how usability testing has improved your projects?
A: Yes, I am a huge fan of simple usability testing, as outlined in the book “Don’t make me think” — this is a short read and I highly recommend it to anyone involved in software design or development. I did a 5-minute Ignite talk on the same topic which you can see here. Usability has been absolutely fundamental to the work we’ve been doing with myWorld for the past five years, to make complex geospatial enterprise data accessible to the average person with little or no training. We believe that the growth in enterprise geospatial applications is all about serving the 95% of people in our customer organizations who don’t know what a GIS is and shouldn’t need to know.
Q: Influenced by some of your usability notes, I once held on to the belief that it was important to signal to users that something in a web map was “clickable” by changing the mouse cursor to a pointer. But now, with touch screens eliminating the cursor, this seems much less important. Has this shift been revealed in your usability testing? And if so, is there a new “touch screen friendly” way to tell the user that something is “clickable”?
A: Your first example is an interesting one actually. In myWorld we are typically dealing with very dense utility maps with many items on the screen. We can easily have several thousand selectable features on the screen at a time. These are typically pre-rendered into raster tiles for high drawing performance, which works very well. But of course that doesn’t give you sufficient information to change the mouse cursor when you are over something that is selectable. We could create UTFgrids to do this, but that’s a lot of extra work with our data volumes, and the density of clickable data on the screen is so great that arguably it doesn’t add much value and could impact performance — you would be trying to rapidly change the cursor all the time as it moved over the map. So in our web application we elected to use a fixed “pointing finger” icon instead of a “panning hand” icon over our map, to give users a clue that they can click anywhere on the map. Depending on where they click, they could get no features or multiple features. But we’ve found that works just fine with a wide range of users who are not technical and have had minimal or no training on the system. They also find panning intuitive even though we the icon we use is different.
Of course on a tablet or phone there is no cursor icon, and users still find it very intuitive and realize that they can drag to pan, click to select or pinch to zoom. So in this particular example I think it’s the case that you don’t really need the feedback of a changing cursor for those operations to be obvious and intuitive.
In the more general case, as web applications or native apps based on them are used more and more on touch devices, I think it’s important to design your application so that it doesn’t need mouseover behavior, which in general is very doable. There are quite a few specific suggestions if you google around, for example this discussion on Stack Exchange.
Q: Is there anything else you’d like to share with the GeoHipster readers?
A: I guess one other topic of the moment that we haven’t touched on is vector tiles. There’s lots of interesting work going on in this area at the moment. I do find it ironic though that when I started doing GIS in the 1980s, most systems were based on tiled vector files — systems were designed that way to get over performance constraints. A huge focus in the industry was getting away from tiles to “continuous” vector based systems — having to split line and polygon features into multiple pieces to fit into different tiles caused all sorts of problems, especially for data editing and analysis. All these problems are exactly the same with vector tiles. So people have laughed at me for saying this, but I strongly believe that vector tiles are not “the future of web mapping” which is the message that I hear from a number of people at the moment. Yes they are interesting and you can do some cool things with them, but they have significant drawbacks too. I believe they’ll be a short term transitory phase, and the real winner of the next iteration will be whoever figures out how to handle a continuous non-tiled vector model in the browser, efficiently loading and unloading features as needed.
I also think that in all the excitement over vector tiles, a lot of people underestimate the strengths of good old raster tiles. For the sort of applications that we do, with dense vector maps and very complex display styles, pre-rendered raster tiles have huge advantages in terms of performance, scalability and portability — the ability to work well on low-end devices, and look the same in all environments, etc. I think there’s a good chance that raster tiles will outlive vector tiles.
Q: Any final words?
A: If you’re not using open source as part of your geospatial applications, you should really take a look at what’s out there. After spending 20 odd years working in the closed source geospatial world, I’ve been really impressed over the last eight years or so both with the innovation going on in the open source world, and with how well the products we’ve been using in our solutions have held up in very large enterprise projects.
Oh, and one more last thing: Mapwheel. I think all geohipsters should have a mapwheel!