I read here that EIP-86 can allow transactions to be paid for by a contract being called. I understand how child-pays-for-parent works in Bitcoin, but EIP-86 does not appear to work at all in the same fashion. I feel like I could spend the time trying to grok the EIP and sort things out rather than skimming it like I just did, but there are a number of things that are unclear to me at first blush. For example, I'm not familiar with Serpent, so I don't know if the example account contract is an entire contract or an entry point for a method -- and, if so, which one; then there's the "(2) and (3) introduce a feature similar to bitcoin's P2SH, allowing users to send funds to addresses" -- but no (2) or (3) near by; I guess that refers to "ring signature mixers" and "custom cryptography".
Could someone please give a high level overview of how EIP-86 enables the contract to pay for the transaction so that I can go back and follow along in the original text? What I'm worried about is DoS attack models (the last part about miner strategies alludes to possible attacks) and would like to be able to assess if that's a real problem. No, I'm not trying to attack it, but I'm also not so knowledgeable that I think I can contribute anything of value to the EIP discussion :) Thanks!