A few years ago, Larry Wall decided to dump Perl 4 for good and to create Perl 5, which was not backward compatible. Right now, the lack of backwards compatibility in Perl 6 is rationalized on similar notes. The situation, however, is completely different.
Perl 4 was a system and scripting language that still originated from the same Perl 1 codebase. It did not have half the features that Perl 5 has, and it was very hard to write maintainable code in it. I've seen Perl 4-like code written in Perl 5, and it was not a pretty sight. (I ended up Perl5-ing it). Perl 4 was an ad-hoc language used by a (relatively to nowadays small) number of UNIX sys-admins and users and was not much more than an awk with an attitude.
Furthermore, Perl 4 code could have been converted with some effort to Perl 5. Perl 5, in fact, still includes some operators only for backwards compatibility to Perl 4.
The situation today is very different. Perl 5 is widely deployed and used by many more users around the world. It is not only used for system scripts, but for automating web processes (in thousands of sites used by millions of online users), for complex text processing, database access, GUI programming (take the Mandrake system utilities and installer for example), games (Frozen-Bubble is a notable example), numerical processing (PerlDL) and almost everything else under the Sun that does not require very fast low-level processing.
Furthermore, it is very usable and supports most of the modern programming paradigms very well: Object-Oriented Programming, Functional Programming, Exceptions, text-processing, garbage collection, nested data structures, arbitrarily-sized data, etc. I'm not saying it's a perfect language, but inventing something entirely different, just to solve its deficiencies is sure to make many users (if not most) unhappy.
Now is the point for incremental renovation. Don't throw everything away. Start with what you have and build on there, without breaking everything. That's exactly what Rindolf aims to be.