On 12/11/2016 15:54, Michael Richardson wrote: > > Brian E Carpenter <[email protected]> wrote: > > registrars. In this case it will be possible for autonomic nodes that > > wish to join the AN to use GRASP with no need for mDNS. If we don't do > > Please be clear: are these enrolled nodes, or new pledges?
Pledges. > > > this, the Anima solution will not be complete in itself. Especially > > during the early formation of an AN after a cold start, this is a > > problem. > > Please define "cold start", here. Power off on all racks; humans pressing factory reset buttons. Systems coming back in random order. All dependencies broken. > > So specifically I want bootstrapping-keyinfra to specify that - > > registrars and proxies MUST be disoverable by both mDNS/DNS-SD and > > GRASP. - pledges MUST use one of these two mechanisms > > Sure, but as a pledge, I don't want to announce a TCP socket that is open for > attack. This is the part that I'm concerned about. The mDNS announcements > From the JA can be multicast, and therefore can be observed without announcing > oneself... Yes, and that's why I thought that in Berlin we'd agreed on a Flood based model (where the proxy floods out its info and any pledge can listen). That's what's coded in demo form as Reggie.py, Procksy.py and Pledji.py at https://www.cs.auckland.ac.nz/~brian/graspy/brski/ . I also coded the discovery approach but it's more complex as well as more vulnerable. Brian > > > Unfortunately I will not be in Seoul to argue this point, so it would > > be nice to settle it by email. > > okay. > > > > > _______________________________________________ > Anima mailing list > [email protected] > https://www.ietf.org/mailman/listinfo/anima > _______________________________________________ Anima mailing list [email protected] https://www.ietf.org/mailman/listinfo/anima
