Surfing the Web Over Wireless

A Core Competence Industry Report
by Lisa Phifer

Wander through any airport today, and you're likely to pass dozens of fellow travelers with cell phones glued to their ears, as if growing strange space-age appendages. Look into any gate area or airport lounge, and you can't miss the crowd of road warriors with laptops propped open upon briefcase or bended knee. After boarding your flight, observe the "air phone" built into your seatback or armrest. As your aircraft passes through 10,000 feet, listen to the hum of disk drives spinning up and the familiar "world of windows" jingle as passengers resume computing for the duration of your flight.

Increasingly, we live in a world obsessed with non-stop communication and computing. Once considered a luxury for upper management, laptops have become so commonplace in corporate America that the purchase of laptops is now increasing at a rate double that of desktop PCs. Cellular telephone and pager usage is growing at least as fast. The number of cellular subscribers in the US increased from 13 to 37 million between 1993 and 1996, tracking an upward exponential curve projected to continue well into the next millennium.

It should come as little surprise that entrepreneurs the world over are staking their claim on the intersection between these two mega-trends: wireless network computing.

According to Information Week, the US mobile data networking market boasts 1.9 million subscribers today, with a predicted annual growth rate of 45% per year over the next five years. What characteristics must a product posses to succeed in this emerging market?

Of course, we expect "smaller, lighter, faster, cheaper" from all evolving technologies -- and wireless network computing is no exception. As a veteran user of leading-edge computing and communications technologies, let me suggest a few additional characteristics: reliable, integrated, ubiquitous, interoperable, and, above all, smarter.

Let's take a look at recent developments in the wireless network computing arena and consider how they fare when assessed against these basic criteria…

From Luggables To Palmtops To SmartPhones

Computers first hit the road more than a decade ago as honking big suitcase-sized "luggables". They've shrunk considerably since that time, to 4 lb. subnotebooks and even palmtops the size of a checkbook and weighing merely ounces. A new crop of hand-held PCs (HPCs) and personal digital assistants (PDAs) have arrived, selling for well under $1000, even dropping under $500. Those running Microsoft's skinnied down Windows CE operating system have fared well: Sharp has sold over 1 million of its Zaurus PDAs, and Casio has been unable to meet consumer demand producing over 30,000 Casseopeia units per month. Perhaps the biggest winner in this emerging market has been the PalmPilot from USR/3Com.

Portable processor speed, storage, and memory have increased in parallel with desktop technology. In 1996, laptops went multi-media, shipping complete with integrated CD-ROM, speakers, and SVGA video jacks. In 1997, laptops, HPCs, and palmtops went mobile: Motorola estimates that between 40 and 60% of all notebooks shipped in 1H97 contained embedded communications devices. The remainder are equipped with a PCMCIA (PC card) slot, easily filled with one of the many after-market wireless modems selling for as little as $300. Entrants in this field include US Robotics AllPoints Wireless PC card, Motorola's Personal Messenger wireless modem, Option's PC Card modem, and Metricom's Ricochet modem.

At the other end of the spectrum, we have analog cellular phones -- also getting smaller, lighter, and smarter. So smart, in fact, that they've begun to creep into data networks to support "thin client" applications such as email and text-based browsers. The Nokia/HP OmniGo 700 is a European GSM cellular phone and personal organizer that can send email and access web pages in Tagged Text Markup Language (TTML) format. Matsushita's Pinocchio and Toshiba's Genio are two Japanese PHS (Personal Handy Phone System) memo pads which include integrated telephones and fax/data modems. The Philips Velo is a full-fledge HPC that can be integrated with a Philips mobile phone. AT&T's PocketNet is a Cellular Digital Packet Data (CDPD) phone equipped with a small LCD screen that can be used to access email, on-line services, and web pages in Hand-Held Device Markup Language (HDML) format.

Laptops, HPCs, PDAs, and smartphones provide a hardware and operating system foundation upon which wireless data networks and applications build. These products are clearly smaller, faster, lighter, cheaper, and smarter than their predecessors, and arguably more reliable. Continued integration of voice, data, fax, and paging services is essential for products that hope to achieve large scale market success. Nobody wants to carry around a palmtop and a cell phone and a pager… And don't forget that other industry mega-trend: multi-media. The International Telecommunications Union (ITU) is working on a variation of the H.324 standard for video conferencing systems to support wireless video phones.

As for interoperability and ubiquity, we must bring our focus up a notch to look at the wireless data network services supported by these devices.

Wireless Workgroups, Wireless LANs, Wireless Data Networks

 USR's AllPoints, Motorola's Personal Messenger, Option's PC Card, and Metricom's Ricochet may all be wireless PC modems, but each of these products uses a different physical and link layer technology.

  • "Plain old cellular", known as Advanced Mobile Phone Service, or AMPS, can be used to support data traffic in much the same way as the "plain old telephone service" in your home can be used to access the Internet with your trusty 28.8 Kbps analog modem. Cellular modems like Option's PC card use modulation/demodulation to transmit circuit-switched data over radio to a mobile base station, across the public switched telephone network (PSTN), to another modem. Data loss and propagation delay over AMPS is high and variable, impeding reliability and reducing effective throughput. As the US cellular market goes digital (D-AMPS), effective data transfer rates should increase due to improved resilience to RF noise and hand-off delay.

  • Another cellular technology, Cellular Digital Packet Data (CDPD), uses spare radio channels in the AMPS or D-AMPS spectrum to carry packetized data (IP packets) to a mobile base station at 19.2 Kbps. Packets are then routed across the good old Internet or other IP network to the destination. By using datagram packets, CDPD-based applications can better adjust to packet loss and delay than their circuit-switched cellular data counterparts. The AT&T PocketNet is based on CDPD data network access, in part because this technology is well-suited for intermittent, low-volume traffic which does not require reliable delivery. CDPD providers charge by packet, not connect time, and utilize data encryption for security.

  • National wireless network providers such as RAM Mobile Data also transmit digital packet data over radio networks. RAM relies on a Mobitex network which offers 8Kbps. These wireless network providers offer store-and-forward messaging to a destination, such as an Internet/Intranet gateway or application level proxy. PC card link level drivers must be matched to the wireless network provider. For example, the USR AllPoints Wireless PC card operates over RAM Mobile Data networks using a special NDIS driver. Like CDPD, these wireless network providers charge according to usage (Kbytes transferred). A variety of Wireless Internet Service Providers offer Internet access over RAM, including GoAmerica, Wynd Communication, Radiomail, and DTS Wireless.

  • Metricom is one of several emergent wireless LAN providers attempting to provide a faster, lower cost alternative to nationwide wireless network services. Metricom uses a combination of pole top radios and wired access points, strategically located throughput a relatively small area such as a campus or urban business center. With effective throughput rates of 28.8 Kbps and flat-rate pricing as low as $19.95 per month for students (plus modem rental), Metricom has been successfully deployed in San Francisco, Seattle, Washington DC, in several major airports, and many college campuses. Metricom's technology is proprietary, and requires use of Metricom's Ricochet modem. A gateway must be used to obtain Internet access through Metricom's network.

  • The list of wireless networks services is seemingly endless. Other Radio-based LAN products can be used to construct private wireless workgroups and campus networks, reaching speeds of up to 2 Mbps. Satellite Networks are both available and planned to provide wireless connectivity, nationally and internationally -- for example, Craig McCaw's Teledesic plans to build a global network of low-Earth-orbit satellites. Personal Communication Services (PCS) intends to link analog and digital cellular networks, packet data networks, and public switched telephone networks into a single, faster, reliable global network. Limited PCS deployment began in late 1996, and providers hope that consumer demand for new services (especially new data services) will drive a broad-scale PCS rollout over the next 5-10 years.

Circuit-switched cellular, CDPD, RAM, and Metricom wireless networks provide physical and data link layer connectivity upon which wireless applications build. Most of these products crawl in comparison to 10 Mbps Ethernet LANs, and are still only half as fast as analog dial Internet access. Wireless modems used with these services are fairly small and light (PCMCIA form factor) -- the inevitable battery accounts for most of the bulk and weight.

A wireless modem is just a hair more expensive than an analog PCMCIA modem, and on par with ISDN and combo LAN/WAN PCMCIA cards. But the true cost issue for most subscribers will be metered usage charges. In a corporate America obsessed with down-sizing and cost-cutting, it may be difficult to budget or justify charges racked up per Kbyte of data transferred. Flat-rate pricing is a significant edge capitalized upon by Metricom.

Today's wireless LAN and wireless data network services get low marks for interoperability. Notable exceptions are CDPD and RAM Mobile Data, both based on standard network protocols. A proprietary wireless infrastructure may be cheaper or faster; weigh these against the more limited options available to you as a subscriber. If you buy a CDPD modem, you can change service providers. The same cannot be said of buying a Ricochet modem.

National wireless network providers offered ubiquitous service -- almost. Wireless communication, by its very nature, is prone to service disruption as you roam about. Wireless network providers offer coverage maps -- read them. Wireless LAN services are constrained to a local area and obviously do not support nationwide roaming. But then, there are many vertical applications of wireless computing which do not require nationwide roaming -- campus nets, mobile workforces, restaurant order entry, hospitals, fire fighters, policemen. Even the guy at the rental car return who prints your receipt. In this case, ubiquitous should mean reliably accessible within the local area of interest.

Middleware, Custom Protocols, and Proxies

Early attempts at wireless computing used TCP/IP directly over a wireless data services to offer common applications such as email. The slow speeds and lengthy delays associated with wireless data transfer prevent this solution from working very effectively (or perhaps even at all). Yet, this approach may still be viable for faster, quieter wireless data services -- particularly wireless LANs.

For the rest of us, several alternatives emerged. Among these were:

  • custom data transport protocols, tuned for operation over a wireless data link or network;
  • customized implementations of TCP, tweaked to improve operation over wireless IP; and
  • middleware which insulates applications from wireless network characteristics.

The first alternative has been evaluated for several years with minimal success. Mobile IP addresses only part of the problem -- enabling mobility of IP addresses. A variety of internet drafts defining custom transports have endured numerous incarnations without yielding a deployed standard. The enormous embedded base represented by TCP simply overwhelms the marketability of transport protocols engineered specifically for wireless.

The second alternative (custom tuning of TCP) has been implemented with some success by companies such a PADCOM. However, these products are highly customized TCP implementations. Commercial off-the-shelf TCP implementations cannot be tuned by end users to completely address the issues surrounding effective transmission over wireless. Most COTS TCP products provide only a few knobs to twiddle, even those which expose numerous configurable parameters hide complex algorithms under the covers, and -- most importantly -- the same tuning must be performed at both ends of a TCP connection.

These alternatives fail to provide a ubiquitous solution -- one which is readily available on any wireless laptop, HPC, PDA, or smartphone and on any wireline application server that might be accessed via wireless.

The third alternative -- wireless network computing middleware -- addresses this concern by constraining wireless adaptation to a (smaller) set of devices, known as proxies. Wireless clients interact with proxies using a middleware protocol which provides reliable, efficient delivery over commercial off-the-shelf UDP/IP. The Hand-Held Device Transport Protocol (HDTP) used by AT&T's PocketNet and the Mobile Network Computing Protocol (MNCP) used by GoAmerica and AirBoss are two examples of this approach. No ubiquitous, interoperable industry-wide solution is currently available to solve this problem. In my opinion, a common middleware standard is needed before robust wireless application deployment can become widespread.

Thin Clients, TTML, and HDML

AT&T's PocketNet and many other smart phones offer short ASCII text messaging services suitable for display in a relatively small window. This approach is often referred to as a thin client. Applications involving plain text, such as integrated email, voice mail, and paging, can be readily supported by this approach. Because limited data storage is available on smart phones, large volume data transfers are inappropriate. And since the user interface (display) is so limited, display of graphics is infeasible.

To enable web browsing from smartphones, vendors have devised alternative "web page" formats for use with wireless. The Nokia's Tagged Text Markup Language (TTML) and AT&T's Hand-Held Device Markup Language (HDML) both address this problem. HDML can be used to defined "decks" of "cards" which are displayed on a smart phone's display. HDML has not been mapped to HTML; this format is currently limited to use with only those web servers which offer HDML-encoded cards.

Devices with additional computing power -- laptops, HPCs, and PDAs -- can use commercial off-the-shelf browsers such as Microsoft Internet Explorer or Netscape Navigator. For these devices, there is no need to limit functionality to only that provided to thin clients. These devices can access any HTML web page using wireless services such as those provided by GoAmerica, RadioMail, and Metricom. Users may apply filters to make more effective use of wireless bandwidth. For example, the AirBrowse service offered by GoAmerica allows the user to filter by page length and graphic image.

The Future Of Wireless Computing

Many industry experts claim that web browsing is the "killer app" that will determine the future of wireless computing. Wireless solutions must become fast enough, reliable enough, and cheap enough to support on-demand access to Internet and corporate Intranet information servers, anywhere, anytime. Will plain-text access to HDML or TTML servers be enough to satisfy our culture's growing appetite for information?

Or will middleware evolve into smart agents, capable of searching out information we define to be of interest, aggregating it, and returning it to wireless clients in condensed form? Will the industry's current fascination with "push" technology become a presence in wireless computing as well? These approaches hold promise for both "thin" and "thick" clients, and I expect to see growth of products with these features.

Corporate acceptance of wireless computing is essential to fund market growth. I believe the key factor will be successful integration of wireless into the corporate network. Wireless remote access must become secure and manageable. Wireless devices must offer seamless access to network operating systems (e.g., Novell) and intranet information servers and databases, and must do so without disrupting the corporate backbone network. MIS managers will need proof that investment in wireless as a remote access technology is innovative and cost-effective, not frivolous or foolhardy.

It is increasingly difficult to differentiate between mobile telephony and mobile computing. I expect this trend to continue, and anticipate integrated wireless appliances that will fulfill my every communication need: voice, data, video -- and computing. I don't expect my HPC or PDA to crunch spreadsheets, but I do expect to be able to search and view their contents, and I don't want to carry a laptop around just to do that.

Smaller. Lighter. Faster. Cheaper. Reliable. Integrated. Ubiquitous. Interoperable. Smarter.

We're on our way, surfing the web over wireless. We've only caught the very beginning of this wave; just how far can we ride it?

-- Lisa A. Phifer, January, 1998

For More Information

To read white papers and product reviews located in our Technology Corner

AT&T PocketNet
PCS White Paper 

To read more about the wireless products and services mentioned in this article…

Casio Casseopeia
Matsushita Pinocchio
Metricom Ricochet
Microsoft Windows CE
Motorola Personal Messenger
RAM Mobile Data
Sharp Zaurus
Toshiba Genio
USR (3Com) AllPoints Wireless PC card
Wynd Communication