> -----Original Message----- > From: Federico Giannici [mailto:[EMAIL PROTECTED] > Sent: woensdag 15 november 2006 17:42 > To: Mark > Cc: users@spamassassin.apache.org > Subject: Re: Bayes column 'token' > > > > Well, bayes_mysql.sql does not specify collation; so, like > > you said, the collation will be your MySQL server-set default. And > > searches in MySQL are case-insensitive by default. Might indeed > > perhaps be a good idea to convert to "latin1_bin" or some such. > > There will be any problem if I convert the current data to the new > collation?
I see no indication (or reason) in the code that tokens are to be handled in an case-insensitive manner. The opposite, ere. So, I'm inclined to say that "latin1_bin" collation is better. I don't wanna be responsible for messing up your database, though. :) So I will test this a bit on my Vmware box. > > PRIMARY for `id` and `token` should not have INDEX for `id` > > and `token` added, too. > > I don't understand what you mean. > The couple (id, token) is PRIMARY, not INDEX... > Where exactly is the problem? PRIMARY, like UNIQUE, always implies INDEX, too. So, adding an extra INDEX for `id` and `token` basically gives you a double INDEX for them. - Mark
