The thing is that it comes down to trusting the people updating the database. What you're trying to accomplish is getting a good set of users by making the buy-in cost fairly high (having to import and export a file manually) so as to not let every no-spelling idiot out there click on the submit button.

This, IMO, is not a good solution. The great thing about FreeDB is that it is going to have almost everything you can think of in it, even if some or much of it is slightly wrong. You want to trade that universality off for correctness, which is probably less useful. What you really need is a large userbase, but a technical way to prevent errors.

As it turns out, I have such an idea. The problem with most of the errors in FreeDB is that people simply submit unchecked info. This usually results in typos. Typos are infrequently duplicated. That is, there is only one correct way to spell "Pink Floyd", but there are many ways to misspell it. If, instead of assuming that the first person who enters the data for a title is correct, you gather a few data submissions first, it should be easy for a program to accept the first duplicated set of data. Since it's more likely that two different people will submit the same correct data than submitting identical incorrect data, this ought to work pretty well.

That being said, if you plan to make money off of this idea, I demand a piece of the action. That is, this idea is mine and I reserve the right to make money from it. (He says as if that has some sort of legal binding.)