GROBJ (Generic Referall OBJect) BoF Minutes, by Joe Hildebrand chairs: Sheng Jiang , Dan Wing IETF76, Hiroshima, Japan, Nov 11, 2009 audio archive: ftp://videolab.uoregon.edu/pub/videolab/media/ietf76/ietf76-ch7-wed-afnoon.mp3 Agenda Problem statement Q: Simon Perreault: What is the problem ICE doesn't solve? A(chair): every app has to do it differently A(Dan): IPTV/RTSP isn't SIP, isn't offer/answer so ICE wouldn't work Q(Markus Isomaki): draft for generic ICE, is that this? A: Yes. Requirements Q(Barry Leiba): what will you do differently based on new apps only or existing apps being chosen? what will change in the design? A(speaker): no change Q(Simon): ICE is more than the addresses A(Dan): coming in the next press Q(Stephen Farrell): confidentiality? A(Dan): TLS/IPSec avoids ALGs, but this is to get the ALGs to help Q(Stephen): Is there anything personal in the refs? A(chairs): good question (Hui Deng): it's just the IP address (Dave Thaler): channel security may not meet the requirements, may need to do object-based. they're different, security-wise Generic Referral Objects Q(Joe Hildebrand): is there a requirement that GROs be small? A(Brian): we're layer 3 guys. binary is good, but you can have a web services encoding if you want Q(Barry Leiba): Is there a requirement for a firewall-punching token? A(Brian): hadn't thought of that Jana Iyengar: I like the idea in the general sense. he doesn't think this is enough. What is the GRO going to have that gets him there? Brian: read the draft. Don't know if we're there yet. Simon: Scope? This is a half-ICE. Brian: don't know enough about ICE Dan: Happy eyeballs. Need something generic. No matter what we do, it won't always work. Simon: I want the second half, too. Dan: that might be too much Dave Crocker: Universal Field Theory would be nice. Diagram similar to the world before BGP. What scenarios? You could represent a DNS entry. Needs at least one real-world scenario to be useful, and he wants it to be useful. Brian: if we had this before ICE, would we have used GROBJ for ICE? Jeff Hodges: URIs as encoding? Brian: no consensus among authors Sudesh: what layer? v6 address selection Dan: there's no v6 address selection available Brian: trying to solve the problem at layer 3, but sometimes only the app knows Sudesh/Brian: this is a layer violation Dan: yes. that's why we do it in apps. Alan Ford: don't do complete solution, he needs GRO for something else Hannes Tschofenig: have used ICE-like for several protocols. Problem is NIH, so reuse is hard. So, who would use this? Dan: don't know. no time right now. Andrew: HIP is more widely-deployed than you think. HIP NAT traversal can be melded in? Someone: in the RFC editor queue, please don't mess that up. Markus Isomaki: hard to define the rules. ICE solved this by just trying. Need an example for typical voice application Dan: we need something real to attach this to Questions Should we work on this? Clear consensus yes. Is the problem well-enough defined? probably not. Is the issue solvable? some people think so. Alexey: need to have another BoF Brian: need to write a problem statement. Need more/different authors? Dan: take to the list. second half of ICE might be a requirement. Dave Thaler 3 requirements: - get/construct scope zone id - GRO, a way to communicate it - path selection can be decomposed and worked on separately.