You don't get into a block.

That sounds like a hard fork then.

The activation height threshold is what creates the rules, but they're still valid to older nodes. I didn't say that it's a good experience. I said this could be used to destroy Bitcoin and more evidently show that the miners are in control. Invalidating new txns ≠ "Hard fork" hard fork is simply something that is incompatible with older software versions.

What happens when I try to spend a SegWit address that I don't have the private key for? I won't get into a block, right? On a legacy node, that is a completely valid transaction because the legacy node has no conception of the witness data. Is Segwit a hard fork then?

Legacy transactions are still valid.

That's what I just said, but they will never be mined into a block because they don't have a valid signature according to SegWit. UTXOs in SegWit addresses technically can be spent by anyone using a legacy node. The only reason none of those transactions make it into a block is because the miners enforce valid witness signatures due to version bit 2.

Right, so a transaction that doesn't follow this theoretical fork will get mined. Basically everyone can ignore it. The big bitcoin complex would create a disincentive for users to interact with them.

No. I don't know where you got that impression. I said the exact opposite.

Legacy transactions are still valid and will be confirmed.

Maybe you don't understand the scenario I'm laying out here. A legacy node(version bit 1), NOT a legacy transaction. This node finds a SegWit address with Bitcoin and produces a transaction without a signature. This is a valid transaction on the legacy node, but when broadcast to the network, anyone running version bit 2 or above will reject the transaction for not having a valid witness.