Sorry I missed your forum post, I have not been following the forums lately.
I think you have a bit of confusion about the role the annotation processor plays. An annotation processor is a compiler plug-in. Loading an annotation processor from the project path would be like loading any other Eclipse plug-in from the project path. How would you unload it when done so as to allow editing of the code? You would need to exit and restart Eclipse.
The reason javac gets away with it is that you *do* exit and restart javac with every new compilation. Even there, though, it's not conceptually a particularly clear-headed idea: it is mixing concerns. The annotation processor is not part of the source code, it is part of the compiler.
Also, yes, it is quite expensive to check for processors in the classpath. The classpath includes jar files that contain .class files. You can't just look for source code; a processor is not required to be source code.
Remember that Eclipse is an incremental compiler, not a batch compiler like javac. It tries to *not* load the entire world on every compile.
If you're using the Eclipse compiler (ecj) from the command line, it does use all the same command line options as javac, and you can plug it into Maven.
I can't speak for the Maven integration in Eclipse, because that happened a long time after the annotation processing code was written.
Thanks,
-walter harley (one of the original Eclipse APT authors)
