ORBit2, IDL compiler improvements

[Elliot Lee <sopwith redhat com>]

Wow, I have some time to hack on ORBit now. I made an ORBit2 module in CVS
where I would like to try and reorganize ORBit a bit to clean up the
messes of the past. Some stuff will be rewritten (new libIIOP by
Sebastian, IDL compiler marshalling/demarshalling), other pieces can be
copied almost verbatim (IDL compiler generation of header files &
typecodes, CORBA_any marshalling, the services), and other pieces will
surive with some hacking (most of the ORB core). The reason for making a
new CVS module instead of a branch on the existing one is to allow
starting from a clean slate.

The goals I want to accomplish by the next major release are to make ORBit
a lot nicer on embedded systems, to make it perfectly secure for use on
exposed networks. Practically speaking these mean:
	. Reduce footprint
		. Add multiple (de)marshalling mechanisms to the IDL
		compiler, such as Michael's thing to use the CORBA_any
		marshalling mechanism, and heuristics to let it choose
		the 'best' one for particular situations.
		. Run memprof on ORBit proggies.
	. Make it faster (which is a goal of mine anyways :)
	. Allow conditional compile-time inclusion of features
	. Add bounds checking in the IDL compiler output.
	. Add bounds checking in the orb code that handles incoming data,
	and formally document + audit the untrusted data that gets
	handled.

I also want to add multithreading & thread safety, just because it seems
to be needed by others. Sebastian already has done some work in this area.

If you disagree with any of these ideas, please speak up - these are just
the rantings of a hacker who is itching to turn the ORBit world upside
down once again :)

-- Elliot
The best way to accelerate a Macintosh is at 9.8 meters per second per second.