COMMUNITY

"Alive and Well"

NeXT Research and Development

In these tough economic times, even corporate giants like IBM and Apple have looked to increase profitability by slashing their R&D budgets. But despite NeXT's recent transition to software, R&D remains a top priority at the downsized firm. "The culture of this company is to do great technology," says engineer Paul Hegarty. "Once that culture is in place, it's difficult to pull it out. Conveniently, it also makes business sense."

In keeping with Steve Jobs's assertion at NeXTWORLD Expo that "R&D is alive and well at NeXT," the company is continuing to improve what has become its only product, rather than merely porting NEXTSTEP to as many hardware architectures as possible.

There will always be only one visionary at NeXT, but it is the R&D group in software that is charged with bringing those visions into focus. NeXT's R&D group is headed up by codirectors of engineering, Paul Hegarty and Jean-Marie Hullot. Hullot also has the additional title of chief technical officer.

Product focus
Early versions of NEXTSTEP were most notable for their ability to tame the complexities of a UNIX-based operating system while still providing all of the power. "In the '80s, a lot of emphasis was put on user interface," explains Hullot. "With InterfaceBuilder and the Application Kit, I think we solved the problem. The key for the '90s is [object] distribution and persistency, and we want an elegant solution to that as well."

The research effort at NeXT doesn't exist in a vacuum. Instead of concentrating on pie-in-sky concepts that may never see the light of day, the emphasis is on products that will actually make it to market. "We do some amount of future research, where we investigate technology that may be of benefit to our customers. Then we take the results of that research and we go through the process of making it real," says Hegarty.

"A long-term vision is fine," adds Hullot, "but at the same time, I don't want people working on something that won't arrive for a few years." Because NEXTSTEP is object oriented, the engineers have been able to add new technologies to the OS as they've been completed, without having to do a complete rewrite every time they come up with an improvement.

For all of the advanced technology that has emerged from the NeXT labs, the R&D effort is concentrated on technology that meets a real business need. For the time being, that doesn't include whiz-bang technologies such as handwriting and voice recognition. For example, NEXTIME, the company's forthcoming standard architecture for media integration, has been specifically tuned to support both real-time video conferencing and on-screen movie playback.

NEXTIME originated as a research project on image-compression routines undertaken by Chief Scientist Richard Crandall back in the summer of 1992. The original intent was to come up with a way to add store-and-forward video capabilities to Mail. "When we started on it, one of the big questions was whether we should provide a framework for teleconferencing or something more like QuickTime," recalls Software Engineer Peter Graffagnino, who now heads the NEXTIME team. "Originally, we thought of it as a way to leverage the networking aspects of NEXTSTEP. It soon became apparent that once you have live movies, you're going to want some way of storing them as documents."

With limited resources and a design philosophy that encourages small teams of people working on very specific projects, NeXT has achieved a better track record in delivering new technologies than many of its much larger competitors. Instead of taking a brute-force approach to software development, the R&D team concentrates on putting enabling technologies in place that can be leveraged by others. "We spend our time working on new technologies like Driver Kit," explains Hegarty. "We could have taken five people and had them go off and write 100 UNIX drivers, or we could have had those same five people go off and design an object-oriented driver architecture that made it simple for people to write drivers."

Looking in, looking out
NEXTSTEP is a combination of original work done at NeXT and the integration of complimentary technologies developed outside the company Ð Mach, PostScript, and UNIX, for example, provide the foundation for the object-oriented system software.

Included among the in-house achievements are distributed objects, interapplication communication, and a user interface that is both aesthetically pleasing and easy to use. With NEXTSTEP 3.0, NeXT added RenderMan to the mix, seemlessly integrating it with PostScript so that develop-ers could add 3-D graphics to their applications without having to develop a new printing model. In 3.1, support for Kodak's PhotoCD was added, making it easier than ever to work with high-quality photographic images within NEXTSTEP.

Hullot says NeXT tries to support existing standards wherever they make sense, often improving their usefulness by integrating them at the system level. After the decision was made to employ an object-oriented storage mechanism in a future version of NEXTSTEP, the R&D group evaluated existing object-oriented databases, going so far as to announce at Object World in 1992 its decision to use software from Burlington, Massachusetts-based Object Design (ODI). The rushed announcement proved ill-advised, however, and NeXT has now abandoned its alliance with ODI in favor of its own approach to persistent objects.

Objective strategy
Just as NEXTSTEP's object-oriented technology has allowed other companies to more quickly develop custom applications, it also speeds NeXT's own development efforts; faster prototyping, the capability to subclass existing objects, and new objects that plug right into the existing system-software architecture are just a few of the benefits. As the principal designer of both InterfaceBuilder and Workspace Manager, Hullot sees his role as the person who makes sure that the benefits of object-oriented design accrue to both developers and users alike. If you want a sneak peak at NEXTSTEP's future, says Hullot, you have only to look a little closer at the present version.

His problem-solving approach to R&D results in technologies that can later be extended throughout the system. For example, the distributed-object technology found in NEXTSTEP 3.0 first appeared as background copy-and-move capabilities in Workspace Manager 2.2; the Portable Distributed Objects that will be available near the end of the year is based on current technology in NEXTSTEP 3.0, 3.1, and 3.2. "It's part of the strategy that we started with distributed objects," says Hullot. "Distribution and persistency should work very well together to provide a very powerful client-server solution that is very well integrated with NEXTSTEP."

by Lee Sherman