Many of "the druggies and drunks" are hardly more responsible for their lot than the mentally ill, and the two categories of problems are highly related. Getting out of addiction is not simply a matter of willpower, and there are many factors that predispose people to end up stuck in those habits that are largely outside of their control. You're being too judgmental.
This is really an artificial problem. There's no point in tackling it, when fuel cells circumvent it neatly.
What's sad is not that post by itself, but the moderation it got, which really showcases the sorry state of technical education prevalent so much that even the average moderator at a supposedly technically-savvy place like Slashdot would confuse fantasy with an good idea within the realm of possibility.
Your comment implies we don't have whole numbers for constants for a good reason, beyond just a choice of a number system. Indeed, you acknowledged this in your response to brantondaveperson below, saying you can't have both pi and e as whole numbers in a single numeric system. What this really implies is that pi and e are each nice and compact characteristics of the physical universe that, if the math was actually representative of said physical universe, ought to be representable in that math as something akin to whole numbers. Of course, they are neither. They are, however, representable in the natural number math quite well: they map directly to the finite algorithms that can compute them to a given precision, whereby the precision limit of the computation (dependent on time and storage) has a dual in the physical universe--the constraints imposed by a finite spacial extent and finite energy density mentioned in my post above. Pi and e to arbitrary precision are not properties of any finite portion of the physical universe; only the finitely-encodable algorithms that compute them are.
The question is whether physics exhibits some signature of an incomplete simulation by a concrete machine with characteristics familiar to us.
Yes, and it depends on exactly what is meant by "characteristics familiar to us". If the simulation hypothesis is correct, the host 'machine' in question is more likely to share characteristics with our universe's physics that have to do with the nature of computability in a qualitative sense, rather than merely quantitative (and specifically, scalability and efficiency). I don't find it implausible that the similarity doesn't much extend to the latter (but does to the former); if that is the case, it still may be the case that the simulation is imperfect as proposed in the paper.
NB that these limits directly imply that any finite region of space can be fully simulated by a sufficiently large, (non-deterministic) linear bounded automaton--an abstract computational machine less powerful than a Turing machine.