Saturday, October 31, 2009

Early Indications October 2009: The Exploding Mobile Web

This newsletter is about numbers, specifically these:







We'll take these in turn.

According to Morgan Stanley's Mary Meeker, whose Web 2.0 presentation should be required reading, the iPhone and iTouch surpassed 50 million units shipped in 9 quarters after launch. This would make it the fastest technology adoption, as measured by zero-50 million, in recorded history. Netscape's Navigator reportedly had 38 million downloads in 18 months, but that could include double-and triple-counting. In addition, Netscape's Internet distribution model allowed it a substantial advantage over conventional logistics, while Apple physically moved all those devices. Actually, an even faster technology adoption was a wireline phone feature: as I wrote in 2005, the federal Do Not Call registry surpassed 55 million users in less than a year, but no software or devices changed hands.

The iPhone has spurred a vast ecosystem of software developers. According to App Shopper, more than 103,000 applications for the iPhone have been approved. While about 20,000 are free and the average selling price is $3.25 for paid apps, GPS add-ons from MobileNavigator and TomTom sell for nearly $100. Other top sellers include mobile editions of both conventional (Uno) and electronic (Madden) games. By contrast, the Google Android Marketplace has 10,000 applications, followed by Blackberry App World at 3,000, Nokia's Ovi Store at 660, and Windows Mobile Marketplace with 246 -- less than 3/10s of 1 percent of the leader.

Those applications are helping drive truly staggering demands on bandwidth. Cisco estimates that global mobile bandwidth demand will increase 66 times (!) in the next four years. Based on AT&T's experience, that number is fully believable:
mobile data traffic has increased 4,962% (essentially 50 times) in less than three years. The wide dissatisfaction with iPhone performance is often blamed on AT&T's network, but provisioning that kind of growth would tax any organization.

As much as the iPhone has stressed the cellular network, the picture would be far worse if wi-fi, which is essentially ten times faster, had not picked up so much of the load. According to AdMob, between 40 and 60% of iPhone data transfer occurs over these ad hoc networks, which were not built with a government stimulus package, a spectrum purchase, or a conscious deployment plan. This offloading of bandwidth may explain why Verizon is allowing its new Google phone, the Droid, to connect to wi-fi, but it is not clear under what conditions or with what fees.

That last number, 100,000, approximates the size of the character set to be allowed by ICANN for Internet domain names in character-based (non-Latin) languages. Right now every root server can look up millions of domain names based on 37 characters: the Latin alphabet, 10 digits, and the hyphen. Starting next year, Cyrillic, Arabic, and other character-based languages will begin to be included.

Given how much of the world a) uses character-based language and b) how fast wireless Internet is penetrating the developing world, the implications of character set for handset design will be fascinating to watch: a RIM Blackberry with 6,000 Chinese characters is not in anyone's future, I don't think.

The expansion to character-based languages may have a deeper implication for the mobile Web. The essential structure of semantic programming is based on a subject (Ridley Scott) - predicate (directed) - object (Blade Runner) model. The semantic triple allows data to be handled more flexibly than in relational databases, where relationships need to be known at the time of the schema's creation. Triplesets, by contrast, can be expanded to form graphs (Blade Runner - grossed - $33,000,000, Ridley Scott - directed - Harrison Ford) in order that web data can become queryable (how many Oscar-winning actors did Ridley Scott direct?). As smartphones become the Internet access device for much of the developing world, how will the various semantics of their many languages inform the deeper structure of Web data and data retrieval? Down the road, the non-Latin Web may have implications for Oracle, Amazon, and IBM at the same time that it challenges carriers and device companies.

On the handset front, meanwhile, Apple has the Graffiti experiment to learn from, along with strong developer momentum as it confronts the question of how to reach the next 50 million users, and the next 50 million after that. Motorola, HTC, Samsung, and Nokia, meanwhile, each bring a distinctive package of strengths and weaknesses to the table as they fight for market share in a global contest for hardware supremacy in a new order. Whatever happens, we will be confronted by growth rates the likes of which no manager (or capital market) has ever seen, each with their own raft of unintended consequences.

For more on semantics, see Toby Segaran, Colin Evans, and Jamie Taylor, Programming the Semantic Web (Sebastopol, O'Reilly, 2009).