Of course by reusing only the concept and not the code, you're going to throw out all the little extras that the old codebase had to deal with along the way... gotchas, unknown requirements, bug fixes, etc, etc...

Good point; however, I've found that the time it takes for someone to sculpt the code into its new context could've been used to write it from scratch. Naturally, this depends on the amount of code. Bug fixes, requirements and the like from the old code may or may not pertain to the code's new home. I've seen this happen too many times. People patch together code from different projects and spend weeks trying to get it to work together because there are so many implementation-specific bits floating around. Of course, YMMV.

You'd be surprised how often people don't reuse code the way it should be done (ie a class library) and instead copy and paste.

It may save you some up-front development costs, but it sure as hell doesn't help maintenance costs and maintainability in general. In fact, it might even be worse if you have to keep the copy-and-pasted code in sync across different applications.

Nice theory, but I work in the real world. Writing from scratch would be wonderful, but I have clients that pay me to deliver solutions, not continually improve with each and every line of code I right. We make the effort when we can, of course, but sometimes compromises have to be made.

As for general purpose assemblies, my experience is that it's sometimes much more efficient to just copy and paste instead of building "flea market" assemblies containing a raft of often unrelated routines. Again, there are exceptions and when you've got a large number of related routines it of course makes sense to package them up, but code reuse through libraries still has a place as well (esp. when you're talking about javascript or D/HTML snippets). Additionally, I'd suggest you consider that often routines need to be tweaked to fit the situation, and often it's more economical (and easier to maintain) to go to your code library as a starting point instead of generalizing or extending (and regression testing) your assembly routine.

"right" in my previous post should, of course, be "write". Dammit, it's early here and I destroyed the credibility of a reasonably decent pseudo-rant :(

Maybe you should read what Joel has to say about this sort of thing:

http://www.joelonsoftware.com