In the future, historians of programming languages will draw a line and say: "this is the point that C++ began it's decline". And that point will be prior to today. The party is over for C++, although many of the partygoers are still drinking punch and throwing streamers in the air.
Peter Seibel's blog excerpts comments from the just-released Coders and Work. He lists multiple comments about the C++ language, all of them from detractors.
C++ has had a history of negative comments. It's early history, as a quiet project and before the internet and related twitterness, saw comments about C++ through e-mails and usenet. As people became interested in C++, there were more comments (some positive and some negative) but there was the feeling that C++ was the future and it was the place to go. Negative comments, when made, were either directed to the difficultly of learning a new paradigm (object-oriented programming), the implementation (the compiler and libraries), or the support tools (the IDE and debugger). C++ was the shiny new thing.
The arrival of IBM OS/2 and Microsoft Windows also made C++ attractive. OS/2 and Windows use an event-driven model, and object-oriented programs fare better than procedural programs. Microsoft's support for C++ (among other languages) also made it a "safe" choice.
The novelty of a new programming language is a powerful drug, and C++ was a new language. Managers may have been reluctant to move to it (the risks of unknown territory and longer ramp-up for developers) and some programmers too (charges of larger executables and "inefficient generated code") but eventually we (as an industry) adopted it. The euphoria of the new was replaced with the optimism of the next release: "Yes," we told ourselves, "we're having difficulties, but the problem is in our compiler, or our own expertise. Next year will be better!"
And for a while, the next year was better. And the year after that one was better too, because we were becoming better object-oriented programmers and the compilers were getting better.
But there were those who complained. And those who doubted. And there were those who took action.
Sun introduced Java, another object-oriented programming language. For a while, it held the allure of "the new thing". It had its rough spots (performance, IDE) but we overcame them and newer versions were better. And C++ was no longer the one and only choice for object-oriented programming. (I'm ignoring the earlier languages such as LISP and Scheme. They never entered the mainstream.)
Once we had Java, we could look at C++ in a different light. C++ was not the shining superhero that we desired. He was just another shlub that happened to do some things well. C++ was demoted from "all-wonderful" to "just another tool", much to the delight of the early complainers.
Other languages emerged. Python. Ruby. Objective-C. Haskell. All were object-oriented, but none powerful enough to dislodge C++. The killer (for C++) was Microsoft's C# language. The introduction of C# (and .NET) struck two blows against C++.
First, C# was viewed as a Java clone. Microsoft failed at embracing and extending Java, so they created a direct competitor. By doing so, they gave Java (and its JVM) the stamp of legitimacy.
Second, Microsoft made C# their premier language, demoting C++ below Visual Basic. (Count the number of sample code fragments on the Microsoft web site.) Now Microsoft was saying that C++ wasn't the shiny new thing.
We (in the programming industry) examined our problems with C++, discussed them, debated, them, and arrived at a conclusion: problems have been solved, but the one problem remaining is that C++ is a difficult language. The next version of the compiler will not fix that problem. Nor will more design patterns. Nor will user groups.
The C++ party is over. People are leaving. Not just the folks in Coders at Work, but regular programmers. Companies are finding it hard to hire C++ programmers. Recruiters tell me that C++ programmers want to move on to other things. We as a profession have decided to if not abandon C++, at least give it a smaller role.
Which presents a problem for the owners of C++ systems.
The decision to leave C++ has been made at the programmer level. Programmers want out. Very few college graduates learn C++ (or want to learn it).
But the owners of systems (businessmen and managers) have not made the decision to leave C++. For the most part, they want to keep their (now legacy) applications running. They see nothing wrong with C++, just as they saw nothing wrong with C and FORTRAN and COBOL and dBase V. C++ works for them.
In a bizarre, almost Marxist twist, the workers are leaving owners with the means of production (the compilers and IDEs of C++) and moving on to other tools.
C++ has been elevated to the rank of "elder language", joining COBOL and possibly FORTRAN. From this point on, I expect that the majority of comments on C++ will be negative. We have decided to put it out to pasture, to retire it.
There is too much code written in C++ to simply abandon it. Businesses have to maintain their code. Some open source projects will continue to use it. But it will be used grudgingly, as a concession to practicalities. Linux won't be converted to a new language... but the successor to Linux will use something other than C++.
No comments:
Post a Comment