I'm inclined to go with... let's see, if I could quantify it, it would be something like 65% tradition, 35% evolution. But I feel I must clarify just what I mean by this, because at the same time I'm for 100% evolution, without contradiction.
On the one hand, I don't think people should change something that they don't fully understand; particularly, they shouldn't `correct' something that would work just fine if only they grasped its application properly. The traditional MAs were the result of the fighting experience of real fighters—in karate, people like Matsumura, Itosu, Chotoku Kyan and Choki Motobu were incredible fighters, extremely dangerous, who did a lot of fighting, though for different reasons, apparently. They and their Okinawan compatriots' criterion was ruthless practicality, and the karate that emerged from their collective experience was a mean, brutal and effective fighting system that had to be seriously disguised before it could be taught in schools. The core of that effective system is still there, in the kata, and it's been rediscovered in a major way during the past decade. But that rediscovery has required the efforts of a lot of very sharp, combat-savvy karateka going over the technical content of traditional kata with fresh eyes. The problem with karate over the past century or so hasn't been with the technical content, but with the way people `read' that content, regarding kata as a kind of dance or meditation or whatever, rather than the summary of fighting techniques that they were constructed to be. So my sense is, you want to hold onto the traditional content as long as it takes you to fully understand it; if you want to make changes to suit contemporary conditions, fine, but you do yourself a serious disservice if you change something without first grasping what it really does and how. So this is the reason for the 65/35 split.
But on the other hand, it's also clear that the creators of `traditional karate' weren't themselves traditionalists in any significant sense. There's the famous story of how Matsumura, confronted by the fighting skills of a Chinese sailor in Okinawa who held his own against him—very unusual, apparently!—offered to help the sailor return to China in exchange for instruction in the latter's combat knowledge, resulting in the Chinto kata. Matsumura couldn't have cared less about where that knowledge came from; he saw that it worked and he wanted to master it himself, period. All of these guys were experimenting and innovating; they cared only about what worked. And one could argue that this is the right approach: keep playing with the system and seeing how it could improved; ironically, that innovative and experimental approach—Chinese fighting skills mixed with Okinawan tuite methods and, as Iain Abernethy points out, a substratum of Minamoto samurai bujutsu thinking about the unity of combat motions regardless of the striking weapon—resulted in the `traditional' karate that some people believe should remain in frozen perfection till the end of time. But if the greatest empty-hand fighters of the past thought that utility was the main criterion and that we should always be ready to innovate when the results justify it—and backed up that view with their own track records in violent encounters—then we might do worse than follow suit, eh?
So I think an approach which is both tradition-weighted and fundamentally innovative is the best one to follow. I know it sounds like I'm trying to have it both ways, but I hope what I've said makes it clear that I'm not...
