Even without Republicans signing up as Democrats and vice-versa, the poll still rests on the assumption that people who voluntarily and without solicitation sign up for an online poll are going to vote in the same proportions as voters in actual elections. The bias is obvious even without respondent deception.
Absolutely, the fact that it’s a self-selected sample makes it useless. I was merely pointing out that even Zogby’s claim that the self-selected sample had 38% Democrats, 35% Republicans and 37% Independents (or whatever) is risible.