It's a hot January morning, and a cluster of engineers are hunkered down, busily testing their apps on a 2G Nigerian network. Things are not going well, due to the high latency and the low-end handsets on hand. It's a grounding experience for developers accustomed to multi-core smartphones running on LTE networks.
"We're handling the timeouts gracefully," says Bill Smith, AccuWeather's VP of digital media and technical operations, "the app doesn't just fail and crash." But there are other problems. "Fetching all that content for the base maps; it takes two or three refreshes to get the base maps to come in. That's a bad user experience."
Smith is eager to see how the app performs on an overloaded network in Jakarta, where we'll be in about two hours. How is this possible? Because we're not actually in Jakarta, or Nigeria, but an Ericsson conference room in San Jose, at the first Internet.org hackathon. Internet.org is the consortium established last year by Facebook, Nokia, Samsung, Ericsson, Opera, Qualcomm, and MediaTek to bring the Internet to the two thirds of the planet that isn't already online.
On this day, teams of developers from Facebook, Twitter, Spotify, AccuWeather, Huffington Post, and others like the nonprofit Water.org have gathered to learn how to make their apps work for most people around the world -- places where 2G networks and single core phones with 320 x 480 displays may be the norm. To make that happen, Ericsson handed out phones and SIM cards and set up mobile networks that simulate what a person would find elsewhere in the world. It lets the developers try out networks in Africa or Indonesia without having to jump on a plane. No need to travel; just swap out the SIM card and you're suddenly in another country.
It's a different perspective from the typical Silicon Valley mindset, which is all about speed and specs. But if you want to make smartphone applications that work for most people (not just most Americans), you have to think slow. Smartphones and the Internet are booming in the developing world. By the third quarter of last year, smartphones overtook feature phones. The world is going online -- wireless Internet service grew by 83 percent in China last year, 84 percent in India, and 68 percent in Nigeria, according to Ericsson, which operates some 40 percent of global cellular traffic. But the mobile Internet in the developing world is a fundamentally different beast than the one we typically talk about. That includes everything from the devices people use to to the plans they purchase, and the networks they run them on.
Towards the beginning of 2013, Facebook's Android project manager Ragavan Srinivasan and some other executives from the company were traveling in Nigeria. A few days into the trip, Srinivasan went into a store, bought a local phone and SIM card, and launched the Facebook app. "We fired it up, and we wait… and we wait… It took a really long time. Even simple things like uploading a photo -- things most Facebook users do -- just weren't working," he says. "That was a pretty harsh experience for us. We'd been building an app for users like us. But we were the exception, not the rule."
When Srinivasan got back to Menlo Park, he began a push to get Facebook working well on Android no matter what kind of device and network it hit. He also changed the way they talked about devices. Rather than referring to the kind of $50 handsets that most people have as "low-end Android," Facebook began to reference "typical Android." That shift may seem like a linguistic nuance, but it belies a completely different way of thinking about application development.
The most common smartphone in the world is, according to Ericsson, the K-Touch W 619. It has a single core processor, and a 3.5-inch display with 480 x 320 resolution. When you hear about the next billion people going online, that's the kind of device they're going to use to do it. What's more, they're going to run that phone on networks where traffic moves at kilobytes per second, not megabytes. In Ethiopia, for example, only 23 percent of network traffic can hit download speeds of greater than 1 Mbps; a mere 4 percent can upload at that speed. For bandwidth-hogging apps, this poses multiple problems. The first is that they may just not work. When phones are constantly trying to suck or push data to a network, and encountering problems doing so, it also eats up battery life. That's a much bigger problem than simply needing to recharge your iPhone twice a day.
"Many of the villages out in emerging markets don't even have the electricity grid," explains Ericsson's Sebastian Tolstoy, VP of business development for radio. "If you have poor coverage, it affects battery capacity, and if you can charge your phone only once a week it's a problem definitely."
Sometimes, these problems can be real head-scratchers. When the team from Twitter was testing its app under the high-traffic Indonesia setting, it found that network reliability issues could mean tweets might take five minutes or more to upload. Given that Twitter is often an of-the-moment experience, how should the app handle that situation? "If you're watching a game and you see a goal, and you tweet 'GOAL!,' you don't want to wait five minutes from now," explains Twitter's Igor Razumeiko. "You'd sound like an idiot."
There also are simple economic issues. The prepaid plans millions of people buy in emerging markets can be wiped out with a single photo upload. A 20-cent day pass in Indonesia cited by Ericsson, for example, only comes with 2 MB of data. A 30-day plan, which costs about $2.07, includes just 20 MB. Things are a bit better in Ethiopia, where $8 per month buys you 50 MB of data.
For Facebook, these realities have meant a complete reengineering of its app so that it can handle different use cases depending on a phone's hardware, the network speed, and the operating system it encounters. "If we know you just have a 320 x 480 display, we just send you those size images," explains Srinivasan.
Other considerations come up as well. As smartphones increasingly penetrate rural areas, they're hitting markets where literacy rates are low. That means thinking about apps in terms of symbols that transcend language. "We deal in the language of iconography," says Ezra Nanes, who runs application products for AccuWeather.
Two hours after the first simulation, Ericsson's engineers switch networks, and now we're running on a crowded cell in Jakarta. It poses an entirely different set of problems than the one from Nigeria. While it has solid signal if you can get on, it's intentionally overloaded and most of the developers can't even connect, including the team from AccuWeather.
"It's tough," says Nanes. "If you're on the golden pipe, you're solid, but we're not even loading."
