-guest.c : use xc_domain_getinfo() rather than xc_domain_getinfolist()
	-will avoid a hypercall for every vm
	-requires some gymnastics around a xc_dominfo_t conversion to 
	  xc_domaininfo_t 
		-not too much of a problem, just annoying. see both xen/ and 
		  tools/libxc/ for related functions and flag bitmask details

-guest.c : add module support so things are not so xen-specific

-guest.c : support for zentific tools data
	-add zentific tools struct?

-guest.c : event xs watches to track state changes
	-fire off RPC to update the backend
	-any other local action ( e.g., automated ballooning)
	-NOTE: this fundamentally changes the nature of the daemon from 
		something that merely observes to something that acts upon
		observations

-node.c : consider high-frequency requests to backend for any new jobs
	  aimed at the given node.
	-act upon them???

-TYPES:
	make NODE_STATS a type unto itself and break up the update_node function
		a bit
	also add VIF_STATS DISK_STATS VM_STATS as necessary and in concert with
		similar zrpc changes

(OLD) TODO / FIXME LIST

- (DONE:PARTIAL) fix edge cases where a domU disappears AFTER update_stats()
	has already started. fetching that domain fails because xenstore has 
	pruned it

- fix xmlrpc_check_response return flow

- dmidecode Not Present failsafe usage of 0-uuid for dom0 is stopgap only
	need more uniqueness. use MAC?

- update ssl_send.*
	make sure BIO_read doesnt die. investigate causes. see acm_dom0

- linked list of observed domains? loop through to verify present or absent.
	this fixes edge cases where last observed state was $BLAH (and then
	crash occured or something). could also (alternatively) use last updated
	timestamp in place of this

- add greater authentication support (with handling to reauth if expired)
	x509 client certificates

- zlib compression of outgoing data to reduce overhead
	(roughly 12:1)

- xmlrpc_check_response may need to pad logged messages w/ newline