CBD is the real deal

Software Magazine, July, 1998 by Dan Kara

Will an industry currently enraptured with the concept of component-based development (CBD), be patient enough to allow infatuation to bloom into a long term relationship, or will CBD end up as a technological summer romance? It could be argued that "component development" falls somewhere after the Internet, thin clients, and supply-chain management in a long line of "hot" technologies and techniques that have recently captured the attention of IT managers and the industry at large. In this industry, however, mindshare is a commodity that can be lost as quickly as it was gained.

CBD will eventually be replaced by other technologies, such as server-side Java and 64-bit computing, in the hearts and minds of magazine editors, analysts, and vendors who are constantly on the prowl for the "next big thing." The replacement will come sooner rather than later judging by the column space dedicated to the recent release of the Enterprise JavaBean specification during Sun's JavaOne Conference, as well as the upcoming tidal wave of ink dedicated to the joys of Merced and 64-bit computing. However, while component development might be pushed off the front pages of industry journals by newer technologies, CBD will not join the ranks of video-on-demand, artificial intelligence (AI), and other "should-a, would-a, could-a' technologies. Unlike these "solutions," both a need and a market for component development applications already exist, and the technology is in place and it works.

No Silver Bullets

The greatest threat to the component development movement -- if it can be called that -- could come from too much exposure. More specifically, CBD is doomed flit is positioned, and perceived, as a silver bullet." The term silver bullet, as you may recall, was first put forth by Fred Brooks, author of the now classic Mythical Man-Month, to describe a single Technology or management technique could magically deliver order-of-the magnitude improvement in the production of high-quality software. 4GLs, Case, and object-oriented development Provide examples of technology and Methods that have been positioned as developmental silver bullets over the years. Recently, the term silver bullet has also been applied to any type of technology that promises greatly increased cost savings, profits, and competitiveness to business. Client/server architectures and the Internet come quickly to mind in this respect.

No, component development is not a silver bullet. It is not that the technology and approach are inherently flawed, but simply be cause silver bullets cannot exist. The process of translating business and technical requirements into working software is a very complex undertaking, and getting more so as execution architectures are extended to multiple tiers and across the Internet. For this reason, component development, or any other single developmental technology or methodology, is unlikely to completely overcome all of the difficulties of building complex, high quality systems in a timely manner.

It is both the complexity of capturing ideas -- actually, this is the most difficult part -- and then realizing them as running systems that makes software development fundamentally different from other types of engineering efforts. This is why analogies comparing component development with hardware engineering are flawed. While hardware systems can be quite complicated, even relatively simple software systems can be orders of magnitude more complex. At some level, all software is deterministic (this is why software security measures can never really be completely foolproof), yet with software of even moderate complexity the combination and permutation of options (including interaction with other software systems) make software systems essentially nondeterministic. This is why testing software is so difficult.

While not a silver bullet, component development is the best method for increasing developer productivity, enhancing software quality, and improving the maintainability and modifiability of the resulting software systems. These benefits are the direct consequence of the component development approach: namely, that systems are assembled from pre-built and pre-tested code modules. Some have even argued that the component approach fosters a type of development caste system, with the more technical developers creating components, and the less-skilled assembling them into sophisticated applications they could not have created any other way.

Actually, component development is a poor choice as a silver bullet (a compliment if you consider the attributes of prior silver bullets). Silver bullet solutions are frequently mandated by management (Case, BPR), and their use often excluded, and then precluded, other approaches (4GLs). Many times, strict adherence to a given methodology was necessary (Case OO), or extensive training was required to become proficient in the new approach (Case, OO). In many instances, the silver bullet simply failed to deliver the goods (mainframe Case, early AI, expert systems, and client/server), or was applicable only within a very narrow domain (Ada, expert systems). As described below, component technology and approaches suffer from none of these afflictions, and it is for this reason, and not because it has a temporary hold on the imagination of the industry, that the approach will succeed. Going forward, component technology and techniques will play some part in all application development and software integration efforts.