My previous post claimed that new platforms spur the adoption of new languages. The more I think about it, the more I believe I was wrong.
New platforms don't simply spur the adoption of new languages. They compel the adoption of new languages.
A platform offers a set of capabilities and a set of concepts. Languages are designed around those capabilities and concepts. Change the platform, and you change the capabilities and the concepts, and you need a different language.
For batch processing, COBOL and FORTRAN were acceptable. They didn't work for timeshare systems and they didn't work for microcomputers. Timeshare and micrcomputers were interactive, and they needed a language like BASIC.
Windows and OS/2's Presentation Manager required a language that could handle event-driven processing, and object oriented languages (first C++, later Visual Basic and Java) met that need.
Web applications needed a run-time system that was constantly present. We started web applications with Perl and C++ and quickly learned that the startup time for the programs was costing us performance. Java and C# load their run-time systems independently of the application program, and can keep the run-time in memory, which gives better performance.
Changing languages (and the mindset of the underlying platform) is a significant effort. One does not do it lightly, which is why large organizations tend to use older technology.
But where does this leave functional languages?
From my view, I see no platform that requires the use of functional languages. And without a compelling reason to use functional languages, I expect that we won't. Oh, functional languages won't go away; lots of developers use them. (I myself am a fan, although I tend to use Ruby or Python for my own projects.)
But functional languages won't become the popular languages of the day without a reason. Inertia will keep us with other languages.
At least, until a platform arrives that compels the capabilities of functional languages. That platform might be the "internet of things" although I expect the first versions will use the currently popular languages.
Functional languages offer increased reliability. It may be possible to prove certain programs correct, which will be of interest to government agencies, banks, and anyone in the security field. (Turing proved that we could not prove correct most programs, but I believe that we can prove correct programs that are subject to a set of constraints. Functional languages may offer those constrants.)
I'm not abandoning functional languages. I like what they offer. Yet I recognize that they require an additional level of discipline (much like structured programming and object-oriented programmed required additional discipline) and we will switch only when the benefits are higher than the cost.
No comments:
Post a Comment