Correct me if I am wrong, but the longest valid chain is the accepted chain no?
No if it's behind the last checkpoint for instance.
Yes, but checkpoints don't affect this system(other than the hard checkpoints we manually put in), as the auto-checkpointing is done locally and only stored in memory, it is lost upon closing the client.
That's what I'd like you to explain in depth including why you think it's better than ACP which is a proven solution.
I have modified my post accordingly,
By ACP, I assume you mean feathercoin's solution, I've not read that much about it in detail however from what I know
ACP may be "proven" but its still centralized... all one has to do is take down the checkpoint server(s) and then proceed with the attack.
This is a decentralized defense system that does the same job, arguably better for certain kinds of attacks.
It is centralised to some extent, though optional while enabled by default. Clients may unsubscribe if they wish to. In order to take down the master node(s), the attackers have to find those first as their IPs are not coded in the client. Even if they succeed, a new master node may be set up in a matter of minute.
This system will simply remove(as best it can) all the longest chains that had any 6 blocks in a row that were produced too quickly (contain 6 blocks produced < 10 minutes).(Past block 100K)
Goldcoin's block target of 2 minutes and 60 blocks to retarget assumed. Your white paper says: "No one peer may transmit more than 5 blocks every 10 minutes, regardless of their origin." Do you understand there is quite a big difference between both statements? What do you use for peer identification? What if there are many peers attacking? What if they fake time stamps? What if it's a legit pool like Multipool or Middlecoin? There are many ifs.