Maybe I got it: when written upper-case it indicates a transformation class, so GBDE is a transformation; when written lower-case it indicates a class implementation. Therefore the GBDE set of transformation is performed via the gbde implementation (module).
With regard to geli vs eli, looking at the source code seems that every module lacks the initial GEOM's G, thererfore gmirror is named mirror and geli is named eli.
So it should be that:
- GELI is a transformation class (i.e., a specification)
- geli is the class implementation (i.e., an implementation)
- eli is the module where the above implementation lives (that is eli implements the geli class).
By the way, GEOM is really the shorten for GEOMetry (got from the geom@ mailing list).
Does it sound good?