abstract
introduction
	very brief statement of how MIP works
	problem statement
	solution statement

	goals
		heal Mobile-IP single point of failure problem
		do it so that is interoperate with MIP itself

		consider both physical and network redundancy issues

	explain - how HARP works
	design goals
		no fate sharing
		opaque to MIP
		no new security problems

implementation (messages)
	messages
	configuration (TBD?)

overview
	assumptions therein

	1. relies on unicast routing internally.
		HARP is an IGP
	2. no mods to Mobile-IP in terms of FAs/MNs
	3. HARP co-HAs work in parallel - share state
	4. HARP systems share one IP address (mobile-IP
		gateway address)
		MUST have one disjoint IP address for
		internal addressing

	topological considerations
		external Mobile-IP/IGP point of view
			picture #1
			registration packets/data packets share
				same external routing
			internal HARP failure
		partitioned subnet question
			two possible topos
			partitioned subnet HAs
				advantage in that HAs may be
					"further" apart
				give two possible HOMES for MNs
			same physical link HAs
				must consider possible difficulties
				caused by both ifs answering ARP requests
				solution: use HARP ping so that only
				one system honors ARP at a time

security and redundancy
	fault tolerance
		redundancy enhanced if HARP peers are 
			physically and logically not close.
	message security
		IPSEC ah or MIP-like message digest authentication.
	types of failure
		total HA failure or interface failure dealt with
		router failure to HA may be dealt with 
		point: HAs should be as logically and/or physically
			distant but still IGP connected;
			e.g., more than one router path to them
			different 

related work
	Numerous cases of pairs.  NIS master/slave,
	DNS master/slave.  Similarities are slight though.

conclusion
	goals satisfied
	src available at

acknowledgements

bibliography