Which ISP do you think is the world’s most popular, measured by number of customers? It’s not one of the business-class heavyweights featured in this month’s survey, though they do carry most backbone traffic. Not even consumer king America Online, which held the top spot right up until this summer. With more than 30 million subscribers, the biggest ISP in the world is now NTT DoCoMo, the wireless unit of Japan’s dominant carrier.

Known as ‘i-mode’, DoCoMo’s service contradicts all the familiar Internet dogma. It makes money by charging for content, not through ads or e-commerce. It succeeded in the face of a decade-long economic depression. (The Japanese tech bubble burst back in 1989, and it really is just a coincidence that DoCoMo sounds like dotcom.) Most importantly, i-mode can only be accessed via cell phone.

The service is aimed mainly at consumers, but businesses use it too. An i-mode phone can give a salaryman instant access to his corporate network and e-mail on the road, without even requiring him to boot up his PC. Support for i-mode has also become vital to every Japanese company that wants to do business on the Web. In 2000, cell phones overtook PCs as the dominant Internet access method, proving that the wireless Web is real-at least in Japan.

Elsewhere, it’s a different story. Though America’s big five mobile operators offer services based on the Wireless Application Protocol (WAP), superficially similar to i-mode, their customers are mostly uninterested. Even Europeans, the world’s most enthusiastic cell phone fans, haven’t taken to WAP. Nearly every handset sold in Europe has the technology built in, but few people bother to use it.

What went wrong with WAP, and right with i-mode? Every expert points to something different: i-mode’s reliance on Internet standards rather than a whole new protocol stack, the large number of content providers signed up with DoCoMo, or even cultural differences between Japan and the rest of the world. But the most important reason for i-mode’s success is Japan’s cellular standard, which lets users stay connected to the Web nearly all the time without racking up the hefty per-minute charges faced by their WAP counterparts.

Always-on connections are now making their way to the West’s digital cell phones, and so is i-mode. DoCoMo has aggressively pushed its technology outside Japan, striking deals with operators worldwide. So far the most important is with AT&T Wireless, in which DoCoMo now owns a larger stake than AT&T itself. But this doesn’t mean that WAP is beaten: It’s still supported by an overwhelming majority of cellular operators, who have a new version ready to roll. WAP 2.0 has plundered many of i-mode’s best features, and will be built into most cell phones in 2002. The question is, should it also be built into your network?

Cell Cultures
The i-mode phenomenon is often written off as purely cultural. “Japan is different,” says Jake Sullivan, vice president of technology at 2Roam, a wireless ASP that hosts sites for both i-mode and WAP. “There’s little penetration of the fixed Internet, and it’s very difficult to get a landline.” Others point out that the Japanese spend a lot of time out and about, whereas Americans are more likely to be sitting in cars or watching TV. And Japanese cities are among the world’s most crowded, making it easy for DoCoMo to provide cellular coverage: Its network reaches more than 99 percent of Japan’s surface area and virtually the entire population.

A look at other mobile Internet services suggests that there is something special about Japan. Behind i-mode, the world’s next most successful ISPs are run by J-Phone and KDD, DoCoMo’s competitors. Their rival (and incompatible) J-Sky and EZweb systems, respectively, have each notched up more than 8 million subscribers, the kind of number that non-Japanese operators only see in projections from the most bullish analysts. EZweb is even based on WAP, proving that the protocol can be successful under the right conditions.

However, most of the aforementioned cultural factors that favour i-mode adoption apply equally to many other countries. Take Finland, often held up as the showcase wireless society: Most phones are cellular, and wired telephony is considered distinctly retro. Yet few Finns access the Internet by phone, even though Telecom Finland (now renamed Sonera) deployed the very first WAP service and still wins awards from the WAP community for its innovation. Customers can use their phones to buy from vending machines or access location-based guides-but they don’t.

The main difference between Sonera and DoCoMo isn’t their home markets’ culture, or even WAP’s and i-mode’s networking protocols and markup languages-it’s the cellular systems themselves. Whereas Sonera’s network is based on the Global System for Mobile Communications (GSM), the same standard used throughout Europe and most of the world, Japan has its own unique standard called Personal Digital Cellular (PDC). Though the details of these systems are complex and perhaps of interest only to radio engineers, PDC has one advantage familiar to all data networkers: packet switching.

Thanks to packet switching, users can remain connected to a network without consuming any bandwidth. DoCoMo keeps its i-mode subscribers online continuously (except during voice calls), and only charges them for the data they actually send and receive. GSM operators can’t do this because their circuit-switched networks require that a phone’s entire bandwidth capability be dedicated to a connection. Users are billed accordingly, paying for each minute that they spend online. Some try to save money by repeatedly disconnecting, but this adds a delay of between three and 30 seconds every time the phone reconnects. On an already-slow 9.6Kbit/sec link, that’s too long to wait before a page even begins to load.

Americans shouldn’t ignore GSM, despite its reputation as a European technology. It’s now the most widespread cellular standard in the United States, the favoured choice of three large operators. Cingular and Voicestream both run GSM networks already, and AT&T Wireless is building one. The next most popular system is Code Division Multiple Access (CDMA), which Sprint PCS and Verizon use. Though theoretically more advanced, this suffers from the same problem: It’s circuit-switched, making it very inefficient for WAP, i-mode, or any other services that send data in small bursts.

Most GSM and CDMA operators have begun to adapt their networks for packet data, through upgrades often called “2.5G.” These are seen as an intermediate step to true third-generation (3G) networks, which will be entirely packet-switched, even carrying Voice over

IP (VoIP). Originally planned to enable data rates of 100Kbits/ sec or more, they would make cut-down versions of the Web unnecessary. Instead of i-mode or WAP, mobile users would enjoy similar services to those available through a desktop PC. But like so much in the wireless world, these plans proved to be overly optimistic.

The most heavily promoted 2.5G technology is the General Packet Radio Service (GPRS), an upgrade to GSM already deployed by most of Europe’s network operators (Sonera was the first, completing its upgrade nearly a year ago). The first phones and data cards shipped this summer, offering speeds of about 30Kbits/ sec-a long way from the 171.2Kbits/sec originally touted by vendors. Many in the industry now doubt that GPRS will ever reach anything near its theoretical maximum, as prototype terminals generated such high levels of microwave radiation that they risked bursting into flames.

Standard Error
Though wireless operators now say that mobile multimedia will have to wait for 3G or even 4G, technologies such as GPRS will still have a great impact. “It’s a huge improvement over what we’ve had in the past,” says Boris Fridman, chairman and CEO of Broadbeam, a wireless software company. With packet switching, services such as WAP and i-mode are finally practical outside of Japan.

Unfortunately for WAP, lack of a packetized Physical layer isn’t the only thing holding it back. Though often described as a standard, it’s really a collection of proprietary technologies cobbled together.

Its origins lie in a project at former carrier Omnipoint, which wanted a wireless Internet service but couldn’t choose between four competing proposals. The vendors behind these decided to pool their resources, but Omnipoint had disappeared by the time a compromise system was ready. It was ultimately taken over by Voicestream, ironically the only major operator that doesn’t yet offer WAP.

The fudged ‘standard’ resulted in confusion and incompatibility. WAP 1.0 through 1.2 (the present version) required that every site be tested with every phone, and sometimes rewritten to suit different devices. Two companies have even claimed patents on parts of it: Geoworks for the menu-based user interface, and Openwave for a method of transmitting new content notifications through a narrowband channel (the equivalent of putting a hyperlink in an e-mail).

Even when it worked, WAP’s deviation from the familiar Internet put extra barriers in its way. It replaced the entire TCP/IP stack with new protocols of its own, requiring a special server called a WAP gateway to convert between them . Encrypted traffic had to be decrypted at the gateway, which gave the gateway owner (the mobile operator) tremendous power. Any company using WAP to carry sensitive information had to set up its own private gateway and force all users to dial it directly, or else strike a deal with every mobile operator that customers or employees might go through.

To anyone setting up a WAP site, the most obvious difference was that all sites had to be coded using WAP’s own Wireless Markup Language (WML). WAP gateways didn’t convert between WML and HTML; they dealt only with protocols. Though other conversion tools were available, their output wasn’t something that anyone with a reputation to uphold would post on the Web.

Many sites were coded by hand, and WML is harder to learn than HTML. Despite the name, it’s really more of a programming than a markup language, complete with loops and variables. As an Extensible Markup Language (XML) variant, it’s also less forgiving than HTML of errors and sloppy syntax: Omitting even one tag can mean that an entire page fails to display.

Far from a mass-market Web service, developers found WAP more reminiscent of pre-PC hobbyist computing. File sizes were limited to only 1Kbyte, crippling one of WML’s genuinely useful innovations: the ability to place several pages of ‘cards’ into one file, or ‘deck’. Some WAP browsers even lacked the most rudimentary text-display feature of all: word-wrap. Pages had to be made to fit the smallest possible cell phone, even though some users might view them on a clamshell PDA with a widescreen display.

DoCoMo took the opposite approach: Instead of standardising the proprietary, it used Internet standards in its own service. Web pages are marked up in Compact HTML, a subset of the full language. This makes converting sites from the regular Web easy, and hand-coding even easier. (It’s really just HTML, with the more complicated features taken out.) Pages will even display perfectly well in a regular Web browser.

Like WAP, i-mode uses its own (proprietary) mid-layer protocols instead of TCP/IP, but it retains the Web’s Application layer, HTTP. This means that its gateways are transparent to Secure Sockets Layer (SSL), the security protocol used by most e-commerce sites. Data can be encrypted all the way from a phone to a Web server, with DoCoMo unable to peek inside in the way that WAP operators can. However, many site owners still strike agreements with DoCoMo, because its billing model enables everyone involved to make money.

Too Cheap to Meter?
New Economy pundits talked for years about ‘micropayments,’ small electronic transactions of around one cent. The hope (and hype) was that these would enable Web content providers to charge surfers very small amounts whenever they read a page, eliminating the need for advertisements and allowing literally anyone to go into business as a publisher. The problem has always been that the Web lacks a billing system, and credit card processing fees exceed the value of the transactions.

Phone companies do have billing systems which routinely record calls costing only a few cents. DoCoMo has extended its own billing infrastructure to i-mode, so that customers pay for Web content through their phone bills. Content providers can charge whatever they like per month (forewarning users in advance, of course), with DoCoMo collecting the money for them and taking a 9 percent commission. This percentage is higher than the cut taken by Visa, MasterCard, or even American Express, but there’s no minimum charge. And it’s much lower than the traditional publishing industry, in which authors are very lucky if they get to keep 9 percent.

But DoCoMo’s system isn’t perfect. It’s based on monthly subscriptions, so it’s not quite the “click here and be charged a nickel to read the next page” model that some micropayment advocates want.

Site owners have also complained that DoCoMo acts as a censor, refusing to bill for sites that compete with or criticise its own. There are rumours that it might start charging a fixed fee in addition to the 9 percent, which could put small content providers out of business. But it does permit many sites to operate that otherwise would not: The banner ads that finance a site like Networkmagazine.com simply couldn’t fit on a mobile phone’s small screen.

The rest of i-mode’s billing is also based on service provided, not time spent online. E-mail, Intranet access, and browsing free sites (although some sites charge a subscription fee, there are still plenty of personal home pages and corporate storefronts that don’t) all cost about two cents per kilobyte. DoCoMo itself offers special value-added services at prices ranging from five cents for a new ring tone to 50 cents for a stock trade. The charges are deliberately low, but they mount up to fat profits if each customer makes several transactions a day.

Billing is crucial to DoCoMo’s expansion plans. The other aspects of i-mode’s success are easily copied: packet switching is coming everywhere, and the Internet standards that the system is based on are by definition open to all. There’s nothing to stop other operators from setting up a copycat i-mode-style service, as its two competitors in Japan did.

J-Phone chose Internet protocols and its own HTML-based Multimedia Markup Language (MML), which includes more advanced graphics features than i-mode. KDD was perhaps more clever, deploying WAP but adapting the protocol so that it could access pages written in compact HTML as well as WML. This gave its users access to many i-mode sites, though not the paid-for content. The original i-mode looks rather humble next to these two, but DoCoMo isn’t worried: The network is already overloaded, and it would prefer that users migrate to new 3G services.

Mail Merge
The WAP Forum, the alliance of vendors and carriers that controls WAP, has followed KDD’s lead. WAP 2.0 adds compatibility with more Internet standards, including Compact HTML, SSL, and even TCP/IP (which i-mode does not support). Perhaps most importantly, it will include a way for servers to push data to clients, and a primitive kind of Computer Telephony (CT).

‘Push’ has a bad reputation on the Web, but it’s vital to e-mail. Most mobile operators have already tried to run e-mail over WAP, building interfaces that allow customers to check certain Web-based systems or (more usefully) any POP3 account . All approaches share the same weakness: Users have no way of knowing whether they’ve received a message until they actually dial in. Push would forward e-mails immediately and enable other services, such as real-time calendar updates.

There are some non-WAP systems that allow push, most obviously paging. In the United States, Cingular and AT&T Wireless both provide e-mail and corporate network access over their Mobitex and CDPD systems. These are already packet-switched and offer wider coverage than WAP, but were designed for a low-bandwidth era. There’s no voice, and per-second data throughput is often measured in bits, not kilobits.

The Short Message Service (SMS) is a type of push technology, built into all GSM and some CDMA phones. It’s wildly popular in Europe, even though messages are text-only and restricted to around 100 characters, so America’s mobile operators are now beginning to offer it. Some have even made SMS interoperable with desktop instant messaging, the only Internet service that has grown as fast as i-mode. (The two aren’t really comparable, though, because i-mode’s users are all paying customers, whereas instant messaging services are free.)

DoCoMo has offered Internet-compatible e-mail since the beginning, but this year it ran into problems as spammers and virus writers attacked i-mode. In July 2001, the company cut its charges for downloading e-mail, in response to customer complaints about unsolicited advertising. It even admits that two-thirds of i-mode users found spam so annoying and expensive that they were forced to change their e-mail addresses at least once.

Worse, the combination of spam and CT has led to a dangerous virus epidemic. A malicious user can send a message to an i-mode phone that makes it dial a phone number, often a premium-rate one. The most serious such script was used to orchestrate a Denial of Service (DoS) attack on the police, with thousands of phones simultaneously making hoax calls.

The possibility of abuse isn’t the only reason that some people oppose adding new features to WAP. The mighty GSM Association is worried that so many options could further confuse an already-fragmented system. In June 2001, it published what it called the ‘M-Services Initiative,’ a set of guidelines for GPRS mobile data. This wasn’t an attempt to replace WAP, as some of the media claimed at the time, but an attempt to put it under some kind of control.

M-Services was endorsed by a long list of phone vendors and mobile operators, with WAP’s four putative father companies at the top. They don’t wish to capitalise on every option outlined in the complex WAP 2.0, but rather desire a set of rules to enforce compatibility, both with each other and the wider Internet. And they know that if their own association doesn’t lay down these rules, NTT DoCoMo will.

— www.networkmagazine.com