Everybody knows that online polls are just invitations to manipulation, right?

Well, maybe not everyone.

After hosting last week's GOP presidential debate, CNBC put up a "who the won debate" question on its Web site. To the surprise of no one who has been paying attention, supporters of Ron Paul flooded the site and handed their man an overwhelming victory. CNBC's response? It took down the poll, then whined about what Paul's supporters had done.

In An Open Letter to the Ron Paul Faithful, CNBC.com managing editor Allen Wastler complained that the Pauliacs went and ruined his efforts to take "a quick temperature reading" of audience reaction.

"Congratulations," Wastler wrote. "You folks are obviously well-organized and feel strongly about your candidate and I can't help but admire that. But you also ruined the purpose of the poll. It was no longer an honest 'show of hands' -- it suddenly was a platform for beating the Ron Paul drum. That certainly wasn't our intention and certainly doesn't serve our readers . . . at least those who aren't already in the Ron Paul camp. . . . When a well-organized and committed 'few' can throw the results of a system meant to reflect the sentiments of 'the many,' I get a little worried."

We'd tell Wastler that he invited exactly what he got when he put his poll online. But it seems that we don't have to. In his own Open Letter to Ron Paul Supporters, CNBC chief Washington correspondent John Harwood says that the poll should have stayed up. "If you sponsor an online poll as we did, you accept the results unless you have very good reason to believe something corrupt has occurred -- just as democracies accept results on Election Day at the ballot box without compelling evidence of corruption," he writes. "I have no reason to believe anything corrupt occurred with respect to our poll. To the contrary, I believe the results we measured showing an impressive 75 percent naming Paul reflect the organization and motivation of Paul's adherents. This is precisely what unscientific surveys of this kind are created to measure. . . .

"Highly motivated minorities can and do exert influence out of proportion to their numbers in legislative debates and even in some elections. They most certainly can dominate unscientific online polls. And when they do, we should neither be surprised nor censor the results."