>  I was thinking about this concept and I don't know how it can be relied on
>  either unless you trust that the shut-off moment of the pump is the same
>  every time. There seems to me to be too many variables even with the same
>  pump to be able to rely on it.

Well, you can generally assume that even if there are several
variables, given a large enough sample the variations will average
out. None of the variables are going to skew the outcome that much,
and most of them are additive rather than multiplicative. A +/- 1
liter variation (that's a LOT) in shutoff point at fillup from 1/4
tank will only result in a .25 MPG variation. Granted, you will never
get within the sort of precision seen in the science lab, but +/- .5
mpg accuracy  over the long term would not be an unreasonable
assumption.

--
John Bange
'90 Vanagon - "Geldsauger"
'90 Vanagon GL - "Wiesel"