Status of Software Reuse 579
To build a decent piece of reusable code takes a lot of care. As you say, it requires more time than most managers are willing to spend, even if it can pay for...
I never bought the idea of wholesale source-code reuse. New stuff needs to be layered with some form of isolation from the old code. I reuse enourmous amounts of code daily by including them in scripts; or using libraries. The language issues consistently come back and bite when the isolation is not there. I have visited that trench; and it is not pretty.
Yes, I happen to be a firm advocate of layering and good isolation. And to include language functionality in all configuration and console environments.
They usually do in most applications. There are relatively few large projects where the development of a specific language is part of the project, at least outside research projects. Most adapt...
If I were to use a favourite OO language very little matches the original Simula; and I have a warm heart for the estethics of smalltalk. Java doesn't fill me with much warm feeling, but it seems to get some systems working reliably.
But we have ossified the language debate. All languages are are hard; but the success of Java proves that there is room for something not so heavily optimized.
And here is the grossest fallacy.
You buttume that some external force provide a language, and a language toolchain; and that work is to be done as a monolithic project within this.
Mu.
Unthink this. A large project will need meticulous attention to the toolchain; and would be pretty well of with a few layers. Perhaps you need your own language. One tailored to the application. I have been toying with a design specifically for accounting, where auditing rules and transactions and t-accounts is a part of the language.
Once there is major money riding on it someone will step forward and write a good, optimizing compiler for that language.
You don't have to say "language" to investor. "Powerful scripting capabilities" is what you tell them. It is a sensible way to break up code complexity. You write the core in some language...
Think about the major projects in software that get reverential nods decades afterwards. Think of C as a part of Unix, and PL-1 as part of multics. Think where t(w)enex would have been today if they had followed this approach.
-- mrr
Alt Folklore Computers from Newsgroups