That's because marc.info doesn't supply a Content-Type header (or equivalent <meta> tag) to specify encoding, so the browser must assume it's using some 8-bit one. On most computers in Western Europe, the US and so on, it'll assume Windows-1252.
You have to explicitly specify you want UTF-8 to get it. This would seem to be strange since it's The Encoding now, but assuming 8-bit is necessary for backwards-compatibility, I'm guessing.
It's the same reason we have to include '<!DOCTYPE html5>' when we want the browser to use HTML5. It's the same reason XHTML failed and we had to deal with quirks mode for years before HTML5. Backwards compatibility sucks.
You have to explicitly specify you want UTF-8 to get it. This would seem to be strange since it's The Encoding now, but assuming 8-bit is necessary for backwards-compatibility, I'm guessing.