IETF 95 - Homenet Agenda

Tuesday, April 5th, 2016
1400-1600  Afternoon Session I
Pacifico A 

0. Administrivia (10m)

Blue Sheets
Note taker - 
Jabber relay - 

1. WG Status Update - Chairs (10m)

Updated Drafts:
- draft-ietf-homenet-dncp-12 (AUTH48 - RFC-To-Be 7787)
- draft-ietf-homenet-hncp-10 (AUTH48 - RFC-To-Be 7788)

    Both should come out shortly

- draft-ietf-homenet-routing-consensus-call-01

    Not moving forward.  Filed "for the record".

New Drafts:
- draft-chroboczek-homenet-babel-profile-00
- draft-lemon-homenet-naming-architecture-00

Drafts on hold pending other docs:
- draft-ietf-homenet-hybrid-proxy-zeroconf-02
- draft-ietf-homenet-front-end-naming-delegation-04 (expired)
- draft-ietf-homenet-naming-architecture-dhc-options-03

2. Routing

Babel profile for Homenet - Juliusz Chroboczek (20m)
- draft-chroboczek-homenet-babel-profile
https://www.ietf.org/proceedings/95/slides/slides-95-homenet-1.pdf

Discussion on support IPv4 - currently the spec says SHOULD.

Markus Stenberg(MS): MUST is not in the charter with IPv4, so SHOULD
should be fine

Lee Howard(LH):  MUST say SHOULD, per the charter. 

*Agreement on SHOULD*

MS: REQ5 HNCP generates keying material which can be used to derive
per-other-protocol key material (pre-shared key) => 'instruct' is
straightforward, but I am not sure about MUST

Mikael Abrahamsson(MA): Seems fine to have HNCP instruct Babel to do

MS: all MUSTs here really imply just that a) it must be implemented, and
b) enabled iff HNCP tells it to
   
MA: Fine with wording on permissions, etc. 

LH: Figure out how its done. Not there yet

MA: Interactions with HNCP - Implementation specific?
 
LH: Some commmon way for HNCP to signal. 

JC: two Implementations - one requires reboot, one communicates over
domain socket

LH: communication to babel should be standardized.

MS: HNCP specifies it already, just no impl. details on how
process-process IPC is done.

MA: thought Babel picked it from kernel

JC: Implementation detail

MS: HNCP node == whole router. hnetd has nothing to do with DHCPv6 or
DHCP sourced routes, and instead Babel propagates them directly from
kernel FIB
(linux has no real RIB except in particular routing daemons)

David Lamparter(DL): dhcpv6 some configuration concepts...(xxx must
revisit audio)

MA: agrees with dhcpv6 or some other prefix

MA:  Specifies the metric. Don't want to specify more or in the new WG?

Chairs: WG CfA on the list
Chairs note that if you aspire to become an HNCP expert, this is a great
time to volunteer to co-author this draft.

     
3. Naming Architecture and Service Discovery (60m)
https://www.ietf.org/proceedings/95/slides/slides-95-homenet-2.pdf

Architecture Draft
- draft-lemon-homenet-naming-architecture-00

Ted Lemon(TL) speaking

Tim Chown(TC): similar talk came up in DNSSD
    
Andrew Sullivan (AJS): Given the number of people who (didn't) read the
document, it'd be a good idea to get the overview out of the way before
we start talking about things that are missing

Dave Thaler(DT): Lost on the third bullet (No support for device rename
from homenet mgmt UI), does not need to be addressed

TL: OK

TC: In DNSSD, was slapped for using an English gTLD .home,

TL: Not sure how to address

DNSSEC Slide: 
    
DT: two Homenets, same UID. both have machine named banana.  trying to
resolve banana.homenet in each location.

TL: Not addressed, this is hard. 

DT: Arch for remote resolution. 

TL: global name in that case. 

DT: No solution for ambigious name.

LH: Population of global zones, ISP should delegate to home gateway. 

AJS: Thanks for writing this. Serious gap in the document - if you
globally ambigious name, lots of stuff does not work. Should state in a
single sentence up front.

I18N issue is a big issue. The names in mDNS is 'hidden' in the UI. 

Stuart Cheshire(SC): Makes a distinction what goes over the wire and what
uses see.  The UI layer could show pictures but on the wire is the same
string. 
    
TL: One thing missing in this document is allowed one to name your
homenet. 

MS: DNS-SD legacy browse is typically hidden (one flat list of services)
and normal browse (one with domains) includes domain names also in
service names in UI, typically; ironically enough legacy browse seems
the more supported type

Avri Doria (AD): home/corp/mail still an item of contention.

MA: why don't we have "xn--foobar" style name? 

Chairs: May not want to get into this.

AJS: "xn--" are application aware.
        
4. Security (20m)

Homenet Security Architecture
- Presentation from Ted Lemon

[not presented]