I find it hard to believe any *working* code is bad enough that it can't be refactored. (Maybe Netscape 4. Maybe.) If the code has been maintained at all, it's had bugfixes and tweaks. Any reinvention will need much luck to avoid making the same mistakes.

I certainly wouldn't throw out anything longer than a day's work (say, 100 lines of Perl). If it's not an optimal solution for the specific case, generalize it enough that you can provide the choice for next time. If the code's not good enough for reuse, fix it so it is.

Otherwise, you'll probably run into many of the same traps as the first time.

(I'm not discouraging people from learning. I'm not discouraging alternate solutions. I'm just saying, throwing out working code and starting over from scratch is the last option in my book.)