Patrick Logan's Radio Weblog posts: Trouble Ahead for Programmers.
Trouble Ahead for Programmers: Choose Your Assembly Language Wisely
Java is dead. C# rules.
Or something like that. The message is to stop building software artifacts in Java, and to start building them in C# (Turbo Java) instead.
Or something like that. A recent author gives Java five years "if we don't get our act together".
We should be so lucky.
The lesson I learn from this is not to build software artifacts in either Java or C#. Instead we should think of these languages and their runtimes as assembly languages. Applications should be built at a higher level, in notations that will last beyond the latest Turbo languages.
Ward Cunningham suggests "If we are to have a [program] life cycle, let us make it long. One hundred years cannot be beyond reason." This is preceeded by his observation, "...flexibility is a property that programs still easily loose. And the loss is due to a failure of design."
Using the latest languages is not enough, and not even necessary in many cases. More important is good design.
You couldn't tell this by observing the transition over the last few months from shelves of Java books to shelves of C# and dotNET books. What if we had as many books at Barnes & Noble or Borders that addressed good design, independent of notation?
I am once again programming in Scheme, a 25 year old notation, now ported to the Java and C# "assembly language" platforms, and easily ported to the next ones that come along in five years. See you then.
[Patrick Logan's Radio Weblog]
Patrick is so right commenting on the recent trend from Java to C#. I agree with Ward (as usual) and with Patrick is that very little design is going on here. And I am certainly not (as Ward and Kent Beck) know talking about "Bing Bang" UML design here. I am talking about flexible code done with Extreme Programming concepts like Unit Tests, Maximum Refactoring and such. The trendy language of the week doesn't matter. Design and good engineering principles do. Interview someone today and ask them if they know what cohesion and coupling are. To a large extent. .NET solves the language issue. The big thing in .NET is that the language is irrelevant, right? Want to program in Scheme? You got it. Same model, same BCL. Want tp program in Python or Cobol or Lisp or Smalltalk or 21 others? Got the picture? Now that we have the language picture down and don't have to subscribe to the "one true language" story, I agree with Patrick that design is needed again. Make your software flexible and refactored. Give it Unit Tests. Become merciless in refactoring based on the faith you have in your Unit Tests and the fact that your code is well designed and factored.
10:22:41 AM 
|
