Библиотека сайта rus-linux.net
Why do programmers reinvent wheels? There are many reasons, reaching all the way from the narrowly technical to the psychology of programmers and the economics of the software production system. The damage from the endemic waste of programming time reaches all these levels as well.
Consider the first, formative job experience of J. Random Newbie, a programmer fresh out of college. Let us assume that he (or she) has been taught the value of code reuse and is brimming with youthful zeal to apply it.
Newbie's first project puts him on a team building some large application. Let's say for the sake of example that it's a GUI intended to help end users intelligently construct queries for and navigate through a large database. The project managers have assembled what they deem to be a suitable collection of tools and components, including not merely a development language but many libraries as well.
The libraries are crucial to the project. They package many services — from windowing widgets and network connections on up to entire subsystems like interactive help — that would otherwise require immense quantities of additional coding, with a severe impact on the project's budget and its ship date.
Newbie is a little worried about that ship date. He may lack experience, but he's read Dilbert and heard a few war stories from experienced programmers. He knows management has a tendency to what one might euphemistically call “aggressive” schedules. Perhaps he has read Ed Yourdon's Death March [Yourdon], which as long ago as 1996 noted that a majority of projects are on a time and resource budget at least 50% too tight, and that the trend is for that squeeze to get worse.
But Newbie is bright and energetic. He figures his best chance of succeeding is to learn to use the tools and libraries that have been handed to him as intelligently as possible. He limbers up his typing fingers, hurls himself at the challenge...and enters hell.
Everything takes longer and is more painful than he expects. Beneath the surface gloss of their demo applications, the components he is re-using seem to have edge cases in which they behave unpredictably or destructively — edge cases his code tickles daily. He often finds himself wondering what the library programmers were thinking. He can't tell, because the components are inadequately documented — often by technical writers who aren't programmers and don't think like programmers. And he can't read the source code to learn what it is actually doing, because the libraries are opaque blocks of object code under proprietary licenses.
Newbie has to code increasingly elaborate workarounds for component problems, to the point where the net gain from using the libraries starts to look marginal. The workarounds make his code progressively grubbier. He probably hits a few places where a library simply cannot be made to do something crucially important that is theoretically within its specifications. Sometimes he is sure there is some way to actually make the black box perform, but he can't figure out what it is.
Newbie finds that as he puts more strain on the libraries, his debugging time rises exponentially. His code is bedeviled with crashes and memory leaks that have trace paths leading into the libraries, into code he can't see or modify. He knows most of those trace paths probably lead back out to his code, but without source it is very difficult to trace through the bits he didn't write.
Newbie is growing horribly frustrated. He had heard in college that in industry, a hundred lines of finished code a week is considered good performance. He had laughed then, because he was many times more productive than that on his class projects and the code he wrote for fun. Now it's not funny any more. He is wrestling not merely with his own inexperience but with a cascade of problems created by the carelessness or incompetence of others — problems he can't fix, but can only work around.
The project schedule is slipping. Newbie, who dreamed of being an architect, finds himself a bricklayer trying to build with bricks that won't stack properly and that crumble under load-bearing pressure. But his managers don't want to hear excuses from a novice programmer; complaining too loudly about the poor quality of the components is likely to get him in political trouble with the senior people and managers who selected them. And even if he could win that battle, changing components would be a complicated proposition involving batteries of lawyers peering narrowly at licensing terms.
Unless Newbie is very, very lucky, he is not going to be able to get library bugs fixed within the lifetime of his project. In his saner moments, he may realize that the working code in the libraries doesn't draw his attention the way the bugs and omissions do. He'd love to sit down for a clarifying chat with the component developers; he suspects they can't be the idiots their code sometimes suggests, just programmers like him working within a system that frustrates their attempts to do the right thing. But he can't even find out who they are — and if he could, the software vendor they work for probably wouldn't let them talk to him.
In desperation, Newbie starts making his own bricks — simulating less stable library services with more stable ones and writing his own implementations from scratch. His replacement code, because he has a complete mental model of it that he can refresh by rereading, tends to work relatively well and be easier to debug than the combination of opaque components and workarounds it replaces.
Newbie is learning a lesson; the less he relies on other peoples' code, the more lines of code he can get written. This lesson feeds his ego. Like all young programmers, deep down he thinks he is smarter than anyone else. His experience seems, superficially, to be confirming this. He begins building his own personal toolkit, one better fitted to his hand.
Unfortunately, the roll-your-own reflexes Newbie is acquiring are a short-term local optimization that will cause long-term problems. He may get more lines of code written, but the actual value of what he produces is likely to drop substantially relative to what it would have been if he were doing successful reuse. More code does not equal better code, not when it's written at a lower level and largely devoted to reinventing wheels.
Newbie has at least one more demoralizing experience in store, when he changes jobs. He is likely to discover that he can't take his toolkit with him. If he walks out of the building with code he wrote on company time, his old employers could well regard this as intellectual-property theft. His new employers, knowing this, are not likely to react well if he admits to reusing any of his old code.
Newbie could well find his toolkit is useless even if he can sneak it into the building at his new job. His new employers may use a different set of proprietary tools, languages, and libraries. It is likely he will have to learn a somewhat new set of techniques and reinvent a new set of wheels each time he changes projects.
Thus do programmers have reuse (and other good practices that go with it, like modularity and transparency) systematically conditioned out of them by a combination of technical problems, intellectual-property barriers, politics, and personal ego needs. Multiply J. Random Newbie by a hundred thousand, age him by decades, and have him grow more cynical and more used to the system year by year. There you have the state of much of the software industry, a recipe for enormous waste of time and capital and human skill — even before you factor in vendors' market-control tactics, incompetent management, impossible deadlines, and all the other pressures that make doing good work difficult.
The professional culture that springs from J. Random Newbie's experiences will reflect them in the large. Programming shops will have a ferocious Not Invented Here complex. They will be poisonously ambivalent about code reuse, pushing inadequate but heavily marketed vendor components on their programmers in order to meet schedule crunches, while simultaneously rejecting reuse of the programmers' own tested code. They will churn out huge volumes of ad-hoc, duplicative software produced by programmers who know the results will be garbage but are glumly resigned to never being able to fix anything but their own individual pieces.
The closest equivalent of code reuse to emerge in such a culture will be a dogma that code once paid for can never be thrown away, but must instead be patched and kluged even when all parties know that it would be better to scrap and start anew. The products of this culture will become progressively more bloated and buggy over time even when every individual involved is trying his or her hardest to do good work.