Also when people think of osgEarth it's wrongly thought of (IMO) as providing satellite imagery, that's not really what osgEarth is about - osgEarth is a geospatial and terrain engine; that can take data a number of standards based formats (it uses GDAL). TerraGear could theoretically be modified to create an osgEarth compatible dataset
- It doesn't require to the data to be streamed, it will equally well work with local data
- In my tests because of the LOD management it places a slightly lower load on the graphics cards.
Satellite imagery generally looks good from one angle, but when I tried flying about with the osgEarth build I actually found that it was terrible over urban areas, but quite good in the mountains and over the countryside.
To quote poweroftwo's notes
Compromises – known limitations of initial implementation
* Only a single material / landclass – (runway) is used in this implementation.
* Typical ground texture resolution is lacking compared to the native FlightGear geo typical runways and taxiways.
** Micro-texture (a.k.a. detail texture) could be utilized to improve effective ground texture resolution
* KML loading / rendering issues
** FlightGear render non-textured models poorly
** Currently all KML found in the path are loaded regardless of their ground texture resolution. distance from the ownship’s position. To improve load performance, some type of spatially organized framework along with use of the OSG paging mechanism such as proxy nodes should be implemented.
** During initial load of a textured model, some runtime visual artifacts are produced due to OpenGL graphics conflicts. Quitting FlightGear and rerunning restores the correct state.
* Fog - works with Rembrandt, but does not otherwise
* 3D clouds – The Up vector gets out of kilter over long distances
* Need to load height fields that are nearby and page out others.
* No airport lighting in this implementation.
What I read into Jeff's notes is that there is actually a lot of work to do before osgEarth could be a complete replacement for the current terrain engine. There has been a fair amount of criticism of the current WS2.0 floating about recently - but without anyone actually saying how good the current terrain is. There are problem areas, but overall the terrain is excellent and can be further improved by texturing and adding appropriate shader use.