You convert it to utf8, and everything is double encoded.

it took me a while to figure that out. The old table I was converting wasn't mine.

wait, you are saying that if you put utf8 data in a latin1 table it remains utf8?

as in, the table doesn't try to read it as latin1?

no. It's like this. The table contains Marco² in latin1, but that should be rendered as utf8: Marco² .

it's like displaying a utf-8 page in latin1. You see all the codepage selectors.

so when I have MySQL do an INSERT SELECT, it complies, and changes all the opcodes to their utf-8 versions == double encoding hell.

I start all my tables utf8 ...

I could fix it by doing: INSERT INTO `utf8` (`newfield`) SELECT BINARY `oldfield` FROM `latin1`;

me too, which is why I was converting this old crap in the first place.