This is a question I've asked myself before too. Here's a back-of-envelope calculation which seems to suggest that diffusion alone will never be enough to keep

This is a question I've asked myself before too. Here's a back-of-envelope calculation which seems to suggest that diffusion alone will never be enough to keep someone alive inside a box. I'd be really interested to hear if someone can confirm or refute this. Let's assume CO2 transfer is entirely diffusion-limited, i.e. the worst-case of completely still air. The diffusive flux of some gas-component (per unit area per unit time) J is governed by concentration-gradient (dn/dx) multiplied by a magic diffusion coefficient D, which depends on the molecular properties of the gas. For CO2 in air D = 1.6e-5 m^2/s. Let's assume the gradient is linear so dn/dx = (n_{inside} - n_{outside}) / length. Now n_{outside) = 0.04% * 1 kg/m^3 = 4e-4 kg/m^3. We should decide what level of CO2 we can tolerate, 0.5% should be on the safe side, so n_{inside} = 5e-3 kg/m^3. Length is a bit trickier. If (a) we assume completely stagnant air inside and outside, length should be chosen as the size of the box ~1 m say. However if (b) we assume that inside and outside are pretty well-mixed individually, due by breathing, wind etc., then length should be about the depth of the hole, or thickness of the box-material, say 1 cm = 0.01 m. For case (a) CO2 leaves our box at a rate of about 7e-8 kg/s/m^2, for (b) 7e-6 kg/s/m^2. From the original article, a human breaths .84 kg of O2/day, let's estimate a production of 1kg CO2/day (the extra carbon atom can't be all that significant), or 1e-5 kg/s. To compute the area of hole needed to maintain 0.5% CO2, we just divide 1e-5 by J, to get the area. In case (a) we need an enormous area of 135 m^2, in (b) "only" 1.35 m^2. This seems to suggest that "air-holes" must work (if indeed they work at all) with air-flow. This doesn't necessary require an over- or under-pressure in the box - a suction can be generated by wind passing over a box with holes on both sides - this process would be far more efficient at restoring atmosphere than diffusion. But would be less reliable.