Today's remote access paradigm is one in which the remote user's laptop is tethered to an analog phone jack. This model will remain useful for a considerable number of users, but for a growing class of users, a more highly mobile paradigm is emerging.

Business and societal forces have combined to create a sense of urgency regarding productive time. Wireless telephone and paging services have been demonstrably useful in helping us turn the time we are stuck in traffic, or confined to an airport waiting area or even waiting for a lunch date to materialize into productive time. With a cellular phone, we can return calls, check voice mail, or make ourselves available for teleconferences anywhere, any time. We can not only react but also be proactive in scheduling around unexpected events when we might otherwise feel left at the mercy of other drivers or forces of nature. A strong incentive for cellular consumption is the feeling and accompanying security of having the ability to be in control.

Industries that rely on a mobile work force, such as transportation and field service, require mobile information systems that interwork easily with their back-end systems. These enterprise systems have already demonstrated their ability to streamline corporate operations for non-mobile users. Extending the corporate Intranet to the truck-mounted or the handheld terminal is becoming a competitive imperative.

The availability of easy to use devices such as instant-on laptops and handheld computers.

Recent technology advances now make it possible to produce handheld devices that are as small as certain cellular phones yet as smart and multi-purpose as portable PC's. These devices are very adapted for wireless communications environments, better able to maintain signal strength and intelligently manage power consumption. From a hardware perspective, mobile network computers, hand-held PC's, personal digital assistants, and four-line browser phones are ready to complement and compete against existing remote access alternatives.

The ubiquitous availability of data-capable wireless networks.

Nationwide wireless data networking services such as RAM, ARDIS, CDPD, and the emerging PCS data services are providing a plethora of choices to support narrowband data applications.

A focus on customizing wireless data friendly applications that insure data delivery even in the harsh conditions of wireless.

However, these applications are usually custom-designed to overcome wireless environments characterized by low bandwidth, high delay, "lossy" and error-prone channels. An annoying echo or loud static crackle is annoying for a voice conversation, but is a showstopper for data!

No single, agreed-upon method for delivering data services to mobile client devices exists today. Carriers are traditionally slow to invest substantially in a service infrastructure without standards. Custom solutions can be costly and there is greater risk of obsolescence and consumer attrition to other products and services.

• Conform to an Internet Protocol (IP) network architecture and use existing open standards where applicable
• Minimize communications consumption when low bandwidth, high latency networks are used.
• Minimize power consumption for low bandwidth devices with limited battery life.
• Allow mobile computing devices to operate in a transient disconnected state. An association between a subscriber and a service (application) should persist across changes in physical location.
• Provide subscriber identification and authentication mechanisms that are IP address independent.

An optimum protocol should provide

• adaptive packet sizing and packet size negotiation, to deal with changes in bandwidth and error rates
• error checking and retransmission mechanisms designed to deal with channel impairments encountered in wireless networks (such as selective acknowledgement mechanisms, to optimize bandwidth use even when loss is detected.)
• data compression, to take full advantage of available bandwidth, and to reduce the subscriber's service costs
• suppression or reduction of communications exchanges designed to confirm a persistently connected state (keep-alive and similar messages, sometimes referred to as traffic management or housekeeping exchanges), to conserve battery life and remain resilient in the face of frequent but short loss of connectivity

• Provide sufficient flexibility in the design so that traditional client-server "Internet" applications operate well irrespective of wireless or wireline network characteristics; for example, it should not be necessary to modify existing server implementations to accommodate wireless access.
• Provide the ability to support both content-rich and simple text application services.
• Allow end users to change communications environments with minimum complexity. Many mobile computing devices will connect to different networks, depending on the availability of a LAN, ISDN, POTS, or a wireless service. Optimally, an end user would be able to install the appropriate network interface and begin communicating over a different type of network, with little inconvenience.
• Provide support for mobile client initiated or pull applications as well as server initiated or push applications
• Accommodate subscriber needs for information integrity, confidentiality, and authenticity.

• Applications on a mobile client communicate with a Mobility Server using the Mobile Network Computing Protocol (MNCP) for reliable and efficient delivery
  
  
• The Mobility Server acts as a proxy agent or Application Relay to
  
  
• Applications Servers, which provide public or enterprise mail or unified messaging, public web, Intranet and database access, etc.
  
  

The wireless community has been critical of commercially available TCP stacks. Two oft-cited arguments against the use of commercial TCP's are:

• Commercial TCP's assume certain properties of underlying networks that are inconsistent with those of wireless networks. Some implementations have been designed with manual tuning parameters to improve operation in low delay, high bandwidth and persistently available bandwidth environments.

• Certain TCP "traffic control" exchanges prove costly in terms of bandwidth and power consumption. This "chatty" behavior cannot be turned off in commercial TCP's.

Certain commercial TCP stacks operate poorly or not at all over wireless services, but this is the result of implementation choices, not inherent flaws in TCP.

TCP can be tuned to operate over wireless networks, but are custom TCP's the best choice?

A disadvantage with customized TCP stacks is that TCP by definition runs end to end, from client to server, so both client and server require customization. It is not practical to re-deploy all servers to which a wireless data client may wish to exchange data; rather, it is more realistic to limit upgrade requirements to a smaller set of "proxy" servers which relay requests to the rest of the network (public Internet and enterprise).

Having concluded it is best to "proxy" services through an intermediate agent, why proxy using a custom a TCP stack, which requires mobile network clients to run specialized software, and may not be most efficient?

The practical alternative introduced by MNCP is to build the equivalent of TCP's reliable delivery mechanisms in so-called "application space" (with added optimization for wireless networks), and run this on top of any commercial UDP stack. This approach has the benefit of introducing a "skinny" protocol that compartmentalizes within the wireless environment those reliable delivery functions specific to that environment.

MNC session control provides provide user validation (authentication), application access control, user registration and deregistration for application services, and application message request/response correlation. These features allow applications to seamlessly deal with changes in physical location and connectivity over time.

MNC protocol exchanges are optimized to minimize communications exchanges and thus conserve battery life and bandwidth, reducing subscriber costs. Timers necessary for efficient protocol operation can be adjusted to provide cost-effective operation over different wireless data services, with no impact on TCP operation between the Mobility Server and wireline application servers.

• facilities for automatically re-establishing application state following link loss
• capabilities to restart a transmission other than at the beginning (synchronization during transfer by sender-receiver)
• authentication and access controls that operate at a finer granularity than host (IP) addressing

MNC is a more generally applicable platform than competing "thin-client architecture" alternatives because it doesn't focus exclusively on services appropriate only for smart phones, nor does it rely exclusively on the Handheld Device Markup Language (HDML) to deliver information services to mobile computing devices. With MNC, Web browsing and information push services based on HDML can be supported for thin-clients. But why limit mobile network computers capable of supporting commercial Web browsers like Netscape Navigator and Microsoft Internet Explorer to mere text? With MNC, these MCD's can take advantage of the richer HyperText Markup Language (HTML), Java and ActiveX features. Similarly, while text-based electronic mail can be delivered to a thin client using HDML proxies at an MNC Mobility server, more powerful mobile computing devices can receive complex document and other binary file attachments as well.

For more information regarding MNC by Bellcore, contact Lisa Blitzer at 732-758-2713.