Gilles, I have just read and enjoyed your post of 1st August. The problem we face, defining when a device is conscious, puts me in mind of Bunge's comments on the problems we have understanding the universe as a whole: we have troubnle because we are trying to see from the outside something which does not have an outside.
The brain device is conscious if it thinks it is conscious. Consciousness is not an objective property, and the best we can do from the outside is to give it a Turing test. As Hans said, the sun is conscious if it we view it in the right way - apply a time line, and some spatial dimensions - and ask it how it feels today. This overcomes your objection to comp2, I feel. I'm not sure I understand your last point ('without ANY interaction with external storage of information about its own structure). A significant part of my mind is stored on the Nokia9110 organizer I lost 2 weeks ago, which is why I'n not functioning so well at the moment. It was not only part of my mind but it stored information about its own structure. Dennett describes old people whose trauma at being removed from familiar surroundings, into which they had 'offloaded' certain computational tasks, is similar to someone who has undergone major brain surgery. On another topic, I've had the thought that the (Brandon Carter) idea that WAP determines all constants and laws could also solve the MWI decoherence poblem - i.e. why we don't see superimposed live and dead cats. I've posted an expanded version of my recent little article at http://www.higgo.com/quantum/buddhism.htm