The hard fork / soft fork argument is mostly turning into a word salad here. The thing that matters is simpler: if a client reaches height X and its consensus path says "this otherwise valid block is invalid", then that client is off the network the moment miners produce such a block and the rest of the economy accepts it. Call it a minority soft fork, a failed UASF, a self-own with version bits, whatever. The node does not care what label we put on it. Policy knobs are one thing. Relay preferences, peering, filters, OP_RETURN drama, spam wars, all of that is messy but survivable because it does not necessarily change what block you accept. But once you move into "reject this block" territory, you are no longer just expressing taste. You are making a consensus bet. If almost everyone follows you, history calls it activation. If almost nobody follows you, history calls it a funny GitHub branch with casualties. The SegWit comparison only works up to a point. BIP148 had social pressure, miner game theory, exchanges watching, wallets watching, and a very public standoff. You cannot just cargo-cult the shape of that event and assume the same magic smoke comes out. If Knots has code that will force rejection past some activation condition, then yes, the honest discussion is not "Luke would never" or "miners probably won't." The honest discussion is exactly what Dave is asking for: point at the code, point at the height/condition, and explain what happens when the first non-compliant block lands. Everything else is just noise.
The hard fork / soft fork argument is mostly turning into a word salad here. The thing that matters is simpler: if a client reaches height X and its consensus path says "this otherwise valid block is invalid", then that client is off the network the moment miners produce such a block and the rest of the economy accepts it. Call it a minority soft fork, a failed UASF, a self-own with version bits, whatever. The node does not care what label we put on it. Policy knobs are one thing. Relay preferences, peering, filters, OP_RETURN drama, spam wars, all of that is messy but survivable because it does not necessarily change what block you accept. But once you move into "reject this block" territory, you are no longer just expressing taste. You are making a consensus bet. If almost everyone follows you, history calls it activation. If almost nobody follows you, history calls it a funny GitHub branch with casualties. The SegWit comparison only works up to a point. BIP148 had social pressure, miner game theory, exchanges watching, wallets watching, and a very public standoff. You cannot just cargo-cult the shape of that event and assume the same magic smoke comes out. If Knots has code that will force rejection past some activation condition, then yes, the honest discussion is not "Luke would never" or "miners probably won't." The honest discussion is exactly what Dave is asking for: point at the code, point at the height/condition, and explain what happens when the first non-compliant block lands. Everything else is just noise.

The hard fork / soft fork argument is mostly turning into a word salad here. The thing that matters is simpler: if a client reaches height X and its consensus path says "this otherwise valid block is invalid", then that client is off the network the moment miners produce such a block and the rest of the economy accepts it. Call it a minority soft fork, a failed UASF, a self-own with version bits, whatever. The node does not care what label we put on it. Policy knobs are one thing. Relay preferences, peering, filters, OP_RETURN drama, spam wars, all of that is messy but survivable because it does not necessarily change what block you accept. But once you move into "reject this block" territory, you are no longer just expressing taste. You are making a consensus bet. If almost everyone follows you, history calls it activation. If almost nobody follows you, history calls it a funny GitHub branch with casualties. The SegWit comparison only works up to a point. BIP148 had social pressure, miner game theory, exchanges watching, wallets watching, and a very public standoff. You cannot just cargo-cult the shape of that event and assume the same magic smoke comes out. If Knots has code that will force rejection past some activation condition, then yes, the honest discussion is not "Luke would never" or "miners probably won't." The honest discussion is exactly what Dave is asking for: point at the code, point at the height/condition, and explain what happens when the first non-compliant block lands. Everything else is just noise.

The hard fork / soft fork argument is mostly turning into a word salad here. The thing that matters is simpler: if a client reaches height X and its consensus path says "this otherwise valid block is invalid", then that client is off the network the moment miners produce such a block and the rest of the economy accepts it. Call it a minority soft fork, a failed UASF, a self-own with version bits, whatever. The node does not care what label we put on it. Policy knobs are one thing. Relay preferences, peering, filters, OP_RETURN drama, spam wars, all of that is messy but survivable because it does not necessarily change what block you accept. But once you move into "reject this block" territory, you are no longer just expressing taste. You are making a consensus bet. If almost everyone follows you, history calls it activation. If almost nobody follows you, history calls it a funny GitHub branch with casualties. The SegWit comparison only works up to a point. BIP148 had social pressure, miner game theory, exchanges watching, wallets watching, and a very public standoff. You cannot just cargo-cult the shape of that event and assume the same magic smoke comes out. If Knots has code that will force rejection past some activation condition, then yes, the honest discussion is not "Luke would never" or "miners probably won't." The honest discussion is exactly what Dave is asking for: point at the code, point at the height/condition, and explain what happens when the first non-compliant block lands. Everything else is just noise.