Calculated distance doesn't match Google Earth

I believe that the distances in Google Earth assume the earth is a sphere with radius equal to the equatorial radius. (You should check for yourself by constructing and loading simple kml polylines with coordinates, e.g., 0,0 0,90 and 0,0 90,0.) So expect differences on the order of 1 part in 300 in Google's distances compared to the geodesic distance.

To add insult to injury, Google Earth treats the edges of a polygon differently from a polyline. The documentation claims that the edges of a polygon are lines of constant bearing. This is false; they appear to be straight lines on a plate carree projection.

Finally, a challenge for Google Earth users: draw a polygon which approximately outlines Antarctica.

ADDENDUM

Here's the data I get asking Google Earth Pro (version 7.1.2.2041) to compute the distances along simple (hand generated) polylines. The 90 segment data breaks the 90-degree arc into 1-degree segments. Distances are in meters.

0,0 to 90,0 (quarter meridian)
WGS84 geodesic 10001966
GE 1 segment   10001959
GE 90 segments 10001839

0,0 to 0,90 (quarter equator)
WGS84 geodesic 10018754
GE 1 segment   10018754
GE 90 segments 10018627

Clearly this data is inconsistent with a spherical model of the earth (meridian lengths are shorter that equatorial ones). However the discrepancy between the 1 segment and 90 segment results is unforgivable. Even though Google Earth is a very useful tool, you cannot rely on it for any accurate measurements. I understand that Google is at the mercy of the data providers for the imagery and height information. However, this is a completely different issue -- Google Earth just doesn't have a consistent geometrical model of the reference ellipsoid.

This problem was reported to Google via the "earth private beta" mailing list (which is monitored by some members of the Google Earth team) on 2011-07-05.

As I checked on several distances greater of 10 000 km at sea level, Google Map distance calculator uses Vincenty's formula, but strangely sets a=b=mean volumetric radius = 63710008m, which induly increases the real distance by Vincenty's formulae by a factor of 1.0022 or 1.0024.

Vincenty's real distance = Google given wrong distance / 1.0022 (or /1.00024).

In other words, Google Map distance calculator assumes the Earth to be a perfect sphere with a mean radius, with WGS84, of 6371008 m (for that simplified formulae, see great-circle distance in Wikipedia).