There was a 'discussion' over on Sawmill Creek about router table flatness and the term +/-.002 was invoked by router guru Pat Warner. I posted a comment that flatness should not be described as a +/- tolerance as it adds confusion to the discussion. The same is true for straightness, perpendicularity, parallelism, roundness, cylindricity, and so on. In all cases where one is describing a geometric form tolerance, zero represents perfection and the deviation is always described as a positive value. Why does it matter? Well, when one says +/-.002 do they mean .002 total variation (supposedly either plus or minus) or .004 total (from one extreme to the other)?
As an example, one can describe straightness more accurately by saying a rail is .010 out-of-flat in a concave or convex condition, or, that the rail is .010 OOF in both (snake shaped). Then, there can be a discussion about which condition may be more desirable and when. In the case of a router table I would suspect that being convex, with the area immediately surrounding the bit being the highest spot, would be the most desirable form of deviation. For a rail it probably makes little difference, with one case being better sometimes and the other case being better on another occasion.
I asked Pat to clarify what he meant but so far I did not get a direct answer. He did say that his RT was flat within .002 within a 10 inch circle around his bit, so I am going to interpret that to mean he thinks a flatness within .002 total is acceptable. If that is the case, and he wants to continue to use +/- terminology, then he should at least change his description to +/- .001. How much clearer might it be to just say .002 total variation?