Ron Johnson wrote:
On Thu, 2003-08-28 at 19:18, Colin Watson wrote:On Thu, Aug 28, 2003 at 01:08:56PM -0700, Mark Ferlatte wrote:Ron Johnson said on Thu, Aug 28, 2003 at 02:16:22PM -0500:With tight budgets and tight schedules, I've *never* seen a project rewritten.Rewriting from scratch is dangerous anyway; you exchange all of the bugs you know about for a whole new set of bugs which you haven't discovered yet. Better to improve what you've got. This can result in a totally different codebase after a while, but at least you have been able to test and release along the way.Amen, brother. I spend a fair bit of my time in Debian trying to stop people from rewriting things and getting them to fix existing code instead. It's an uphill struggle: for some people rewriting from scratch seems to be much sexier.
It can (and often is) be easier to rewrite if the original author isn't around to explain the workings, and the language/application-framework doesn't have decent debugging outputs.
I've seen maybe one or two occurrences where rewriting from scratch was actually worth it. Before that's the case, the existing code has to be a complete nightmare and you have to have a deadline to add some new feature or other to it. It does happen, but it is (and should be) rare.Unfortunately, if the fundamental design is Really Flawed, then incremental changes are impossible. Yet I also agree that rewritingswaps old bug for new. There's a Catch-22 sometimes.
It has the advantage that you are the current author, are intimately familiar with the new code, and can easily fix any bugs. If you leave prematurely, it's back to square one.