At the heart of this issue is one of trust - is the community of software developers and users willing to trust MySQL AB and its leadership?
As a sometime member of both communities, I will say that, considering the long- and out-standing contributions that the company has made, my answer is yes.
The reason that Bayesian spam filtering works is because it is personalized. Different users get different kinds of legitimate, non-spam e-mail, and for that reason each person has a different set of values in a Bayesian filter. Centralizing a database using averages from many different clients makes Bayesian filtering several orders of magnitude less effective, not more as you seem to assume.
At the heart of this issue is one of trust - is the community of software developers and users willing to trust MySQL AB and its leadership? As a sometime member of both communities, I will say that, considering the long- and out-standing contributions that the company has made, my answer is yes.
The reason that Bayesian spam filtering works is because it is personalized. Different users get different kinds of legitimate, non-spam e-mail, and for that reason each person has a different set of values in a Bayesian filter. Centralizing a database using averages from many different clients makes Bayesian filtering several orders of magnitude less effective, not more as you seem to assume.