will never be strictly competitive even if it were cost-free to run because p2pool will always have a higher orphan rate vs. well-run centralized pools
Except that, in practice, it has a lower empirical orphan rate. The design currently has a communications advantage.
uses to much resources. If you use a pool, you can run mining software on a raspberry pi or a basic router.
The smallest mining device that I'm aware of that you can currently buy which is within a factor of ten of the best $/GH devices (e.g. remotely competitive at all) costs about $2000. Why is a ~$2 bottom of the end cellphone/stb microprocessor your target device for controlling thousands of dollars of hardware? Doesn't this seem more than a little ridiculous? Especially when the consequence is an abdication of control which undermines the security assumptions of the Bitcoin system and which if exploited could leave your hardware and the Bitcoin previously produced by it worthless?