I spent a couple of hours researching the Google Maps and Google Eart projection question. Found a lot of information, but had to piece together the full story. I'd appreciate any corrections to my investigations.
I am currently using a spatial reference with a EPSG code of 102113 for ArcGIS Server map services that are to support mashups with Google Maps.. This is the original Web Mercator code. This was later changed to 102100 for the Web Mercator Auxiliary Sphere, which was more of a sphere than 102113. Last year 102113 was changed to 3857 and adopted as the official EPSG code. 102113 and 102100 were temporary codes as the official governing body was playing hardball with Google and others -- saying that their projection was not a real projection and pretty poor science and math. Apparently they learned how to play nice in the past 18 months.
EPSG 3857 is supposed to be the best for mashups. EPSG 102113, 102100 and 3857 are supposed to be absolutely identical projections, but the latter is official and the formers are not.
In ArcGIS Server 9.3 map services ESRI published the EPSG code of either 102113 or 102100 depending on which map projection was used. In the release of 10.0 ESRI published the EPSG code of 102113 or 102100 as the official code of 3857. In the release of 9.3.1 and 10.0 SP1 ESRI reversed itself and went back to publishing the 3857 EPSG code as 102113 (the original temporary code). Google Maps recognizes this code in a mashup.
Google Earth uses a WGS84 geographic projection with an ESPG code of 4326. This projection is non-spherical (and thus unlike the Google Maps projection). Esri provides a transformation to move between the two.
Not sure about this one, but from what I can gather if you are creating an OGC WMS service then you need to use the WGS84 World Mercator projection (EPSG code of 3395). This is similar to a pure WGS84 projection, but is a Projected rather than Geographic projection.
Google Earth uses a Perspective projection to display the world. Every time you pan, zoom, or tilt the view, Google Earth creates a new custom coordinate system based on your current observer position and direction of view. It does it really, really fast and might be using technology different than what we think of as doing a map projection in GIS.
It doesn't really make sense to talk about the map projection of a 3-D globe viewer. Map projections are about 2-dimensional maps. The whole idea of a 3-D viewer is to have a virtual experience, where your brain realizes the world is a 3-dimensional object and naturally understands that something on the edge of your view is going to look smaller because its farther away. There's not much temptation to take a screenshot of Google Earth and try to measure the distance from New York to Tokyo, which you might mistakenly try to do with a printed map.
I'm sure there are many ways to abuse Google Earth, but the coordinate system it uses isn't one of the concerns. When you use the ruler in Google Earth to measure a distance, it gives you the true geodetic distance. ArcMap was incapable of doing that before version 10, and you still have to know to ask it for geodetic distances (or work in latitude/longitude coordinates) rather than planar distances. The planar distance could be wildly inaccurate if you are using an inappropriate map projection (such as web Mercator).
The Web Mercator projection is OK to use in web applications where you want to be able to zoom in close to any place in the world and the software understands the change in map scale as you move north and south. Go to maps.google.com, pan from Alaska to South America, and watch the scale bar change. I think web Mercator is a poor choice to use for an application designed to focus on one region the size of Montana. The scale of the web Mercator projection at Plentywood is 8 percent greater than the scale at West Yellowstone, and it makes the state look wrong. The only reason to use the web Mercator projection for Montana applications is the ease of mashing up with all pretty base maps in that projection.
Web Mercator is totally inappropriate for use in GIS applications. If you look at Montana in web Mercator projection in ArcMap 9.3, the measure tool will tell you the length of the international boundary is 830 miles. It actually is only about 540 miles. The land area of Montana in the web Mercator projection is approximately 320,000 square miles. In reality, its only about 145,000. The Montana State Plane coordinate system will give you area and distance measurements that are within one-tenth of a percent of the correct geodetic value, anywhere in the state.
I missed part of your point. If people are doing screen captures of Google Earth or of any web Mercator map application and then performing measurements or heads-up digitizing from that, then this could be very bad for them.
All the more reason to use Montana State Plane coordinates for applications that focus on Montana.
Here's a document that I found useful in understanding the limitations of Web Mercator:
Good find Bryan - Great reference! I like the error curves. That should make someone think about what they are using.
© Montana Association of Geographic Information Professionals
MAGIP is a 501(c)6 non-profit organization in Montana
Full Scope Association and Event Management
1236 North Ave. W, Missoula, MT 59801