I read this article, and it seems to me this guy came to a conclusion before he came to an experiment.

What he DID prove is that a Netflix server in LA was only handing out 50KB/sec per http socket. Most web type servers will do this when under heavy load- better to give everybody a little bit than a few people a lot and the others nothing. I think this is correct behavior for a heavy-load situation.

However, when he accuses them of throttling, along with the way this article is titled, STRONGLY implies that they are throttling specific users who use too much. If he wanted to prove this the test is simple- log out of netflix and log in with a friend's account, preferably a friend who doesn't stream much.

Throttling also implies that Netflix is intentionally reducing the connection quality. I see no logical reason for them to do this to EVERYbody, as that would make the Instant Watch service useless for everybody. Far more likely, as stated above, is that he's on an overloaded server.

So my take on it is this article is incompletely researched, draws a bad conclusion (which doesn't make much sense) from too little evidence, and doesn't perform the one test needed to actually verify it's claim.