Quote from: hv_ on April 04, 2016, 07:04:14 PM

Quote from: gmaxwell on April 04, 2016, 03:55:17 PM

Quote from: hv_ on April 04, 2016, 10:20:17 AM

Just to summ up / that attack would be like that:
A miner
1) modifies his own mining code to enter already the SW link into the correct place
2) mines a BAD block NOW, Long time before SW goes live
3) Now he can do lots of real txs like spending BTC (but get back when)
4) SW goes live and all blocks down to his BAD block get orphaned and his spending from 3) are neglegted
? correct ?

Soft-forks are not enforced on the older parts of the chain. Only in blocks from two weeks after 95% of the hash-power had signaled their intent to enforce (counted over a 2016 block interval after the change start-time).

For prior soft-forks, like strictder, there are many historical violations before where it was enforced. It does not make the chain invalid, only violations after the point where the new rule's enforcement begins would do so.

Thanks & yes, I got that. But this soft fork happens at 4) or two weeks later.
The question is, whether the new coded miners (SW enabled) will orphane the non SW chain down two that BAD block made in 2) because they will all agree that this BAD but old block is the last valid SW one since it contains the correct LAST correct SW entry?

The problem is that the bad block cannot exist in the first place. You cannot spend from an output that you do not own. Segwit does not change this behavior.