Add comment on squares, and formatting. #469
Conversation
| # a 2.8x3.5 meter area. | ||
| # | ||
| # Two encoding algorithms are used. The first 10 characters are pairs of | ||
| # characters, one for latitude and one for longitude, using base 20. Each pair | ||
| # reduces the area of the code by a factor of 400. Only even code lengths are | ||
| # sensible, since an odd-numbered length would have sides in a ratio of 20:1. | ||
| # sensible, since an odd-numbered length would have sides in a ratio of 20:1 | ||
| # (which would not be a square). |
There was a problem hiding this comment.
Noting that the resulting area is not a square doesn't add much here in my opinion.
First, there's the whole issue of all areas not even being rectangles in the first place, because they are lat/long-aligned pieces of sphere surface. Second, even if we accept those as square-ish enough for our purposes, the algorithm will later divide those into 4x5 grids, making longer plus codes less square-ish anyway.
In the end, the rationale for not dividing along one but not the other axis is not that the result is different from a square per se, but the fact that we're interested in minimizing the potential error between the actual location we're putting in and the reference location we're getting back - and having better precision just along one of two axes doesn't help us with that.
No description provided.