IETF
dhc
dhc@jabber.ietf.org
Thursday, 17 November 2011< ^ >
tomek has set the subject to: DHC IETF'81
Room Configuration

GMT+0
[00:54:13] Yuri Nawata joins the room
[01:05:29] Suz joins the room
[01:08:04] fdupont joins the room
[01:08:34] pselkirk joins the room
[01:08:44] Chris Donley joins the room
[01:08:44] kazubu joins the room
[01:13:34] <Chris Donley> need new authors for option-guidelines, dhcpinform-clarify
[01:13:46] <Chris Donley> reconfigure-rebind ready for IESG
[01:14:56] twintersunh joins the room
[01:15:33] <Chris Donley> Class-based-prefix
[01:17:15] <Chris Donley> slide 1. Motivation - scoped prefixes. Need metadata
[01:17:30] <Chris Donley> slide 2
[01:18:34] maximouton@jabber.org joins the room
[01:18:47] maximouton@jabber.org leaves the room
[01:19:21] maximouton@jabber.org joins the room
[01:19:31] <Chris Donley> wlan-epc - proper address selection can offload some mobile traffic
[01:22:04] <Chris Donley> slide 3 home network scenarios
[01:22:38] naito.kengo@jabber.org joins the room
[01:23:06] maximouton@jabber.org leaves the room
[01:23:07] <Chris Donley> slide 4 - proposed extension
[01:23:47] <Chris Donley> prefix class in iana and iapd messages
[01:23:56] maximilien mouton joins the room
[01:24:05] maximilien mouton is now known as maximouton
[01:24:28] maximouton leaves the room
[01:26:15] <Chris Donley> woj - client would use this for source address selection?
[01:27:07] <Chris Donley> woj - what scope do these addresses come from?
[01:27:47] maximilien mouton joins the room
[01:27:55] maximilien mouton is now known as maximouton
[01:27:58] maximouton leaves the room
[01:29:41] <Chris Donley> how would the prefix class be defined?
[01:29:56] <Chris Donley> TBD
[01:29:59] twintersunh leaves the room
[01:31:12] <Chris Donley> wg mailing list
[01:31:31] <Chris Donley> Next - Woj -Client Hardware Address option
[01:32:28] <Chris Donley> DHCPv4->v6 transition
[01:32:52] <Chris Donley> correlate DHCPv4/v6 messages
[01:32:55] tatsuji joins the room
[01:33:29] <Chris Donley> copy dhcpv4 client hw addr option into dhcpv6
[01:34:57] <Chris Donley> jjb - vendor specific option?
[01:35:48] <Chris Donley> we considered this in the past, and decided to use vendor options for more info
[01:36:01] <Chris Donley> Woj - CableLabs option too specific
[01:36:15] <Chris Donley> JJB - could create BBF vendor specific option, too
[01:37:03] <Chris Donley> Ralph - had a reason to do it in DOCSIS, but weren't sure about general applicability, didn't want to wait on WG.
[01:37:30] <Chris Donley> reasonable to ask about general applicability, otherwise vendor specific option might be appropriate
[01:38:36] <Chris Donley> Ted - missing timer, can't be used as identifier
[01:40:27] <Chris Donley> next preso
[01:40:50] <Chris Donley> dhcpv4 relay encapsulation
[01:40:59] TACHIBANA toshio joins the room
[01:41:40] tsavo_work@jabber.org/Meebo joins the room
[01:41:42] <Chris Donley> next slide
[01:42:09] <Chris Donley> next slide
[01:43:08] <Chris Donley> next slide
[01:43:17] <Chris Donley> request review/comments
[01:44:25] <Chris Donley> Ted - interop testing, WGLC soon
[01:44:49] <Chris Donley> next preso - redundancy considerations
[01:46:45] <Chris Donley> updates after WGLC
[01:47:18] <Chris Donley> next steps
[01:47:54] <Chris Donley> ok to informatively reference failover drafts? yes - won't delay publications
[01:47:59] <Chris Donley> submitting to IESG
[01:48:29] <Chris Donley> not dhc wg doc
[01:48:50] <Chris Donley> next preso - dhcpv6 suboptions
[01:49:44] Chris Donley leaves the room
[01:50:05] Chris Donley joins the room
[01:50:37] <Chris Donley> possible solutions
[01:50:52] Chris Donley has set the subject to: DHC IETF'82
[01:54:22] <Chris Donley> didn't reach consensus on mailing list
[01:54:34] <Chris Donley> calling for a vote here
[01:55:13] <Chris Donley> Ted: what statement are we making when we send an ORO? I support/am interested in the following options? I want the server to send the following options?
[01:58:29] tsavo_work@jabber.org/Meebo leaves the room
[01:58:55] <Chris Donley> vote
[01:59:00] <Chris Donley> Option 1: 2 people
[01:59:18] <Chris Donley> Option 1+ Option 2: 3
[01:59:34] maximilien mouton joins the room
[01:59:43] maximilien mouton is now known as maximouton
[01:59:59] maximouton leaves the room
[02:00:08] <Chris Donley> unsure, decide later: 2
[02:00:24] <Chris Donley> Want someone else to decide: several
[02:01:19] <Chris Donley> next preso - route option
[02:01:24] <Chris Donley> not dhc wg item
[02:02:01] <Chris Donley> -03 changes
[02:03:07] <Chris Donley> wglc in MIF
[02:03:50] <Chris Donley> questions/comments
[02:05:30] <Chris Donley> implementations
[02:06:02] <Chris Donley> ISC/Dibbler implementations worked on by 2 different people? yes
[02:06:09] <Chris Donley> DHCPv6 vs. RA
[02:08:03] <Chris Donley> Ralph - If dhcpv6 vs. RA send conflicting info, is this a config error, or are there real-world scenarios?
[02:08:30] <Chris Donley> 2 routes from different providers
[02:09:00] <Chris Donley> Ole - why is this specific to route option?
[02:09:21] <Chris Donley> Ted - Architectural discussion to be held later
[02:09:30] <Chris Donley> Ralph - different WG? Ted - maybe
[02:09:55] <Chris Donley> post MIF comments
[02:11:38] <Chris Donley> JJB - clarifying who needs this - Roberta: Telecom Italia
[02:12:19] <Chris Donley> Roberta - BBF
[02:14:41] <Chris Donley> next steps
[02:17:13] <Chris Donley> Next preso - MaxSolRT
[02:18:15] <Chris Donley> Background - DHCPv6 client sends solicit until it receives a response
[02:18:54] <Chris Donley> rfc 3315 defines retransmission exponential backoff mechanism
[02:19:10] <Chris Donley> sol_max_rt default 120 s, changeable
[02:19:39] <Chris Donley> problem statement - steady state is 1 solicit/client/120s
[02:19:55] <Chris Donley> aggregate traffic can be significant
[02:21:10] <Chris Donley> solution 1: change default sol_max_rt
[02:22:17] <Chris Donley> solution 2 - add shut up and go away option from server to client
[02:22:53] <Chris Donley> next steps - add IA_PD to list of options
[02:23:04] <Chris Donley> choose an option (1, 2, or both)
[02:23:54] <Chris Donley> solution 1 may add 30 min delay after DHCPv6 service restored
[02:24:47] <Chris Donley> solution 2 adds per-client delay time half of Sol_max_rt
[02:26:37] <Chris Donley> JJB - if the retry is too long, some clients will seemingly hang/fail to initialize. User may complain about slow delay.
[02:27:48] <Chris Donley> JJB - minimum requirement to be able to send the option in IA_NA, IA_PD, etc.
[02:28:14] <Chris Donley> Ole - what consequences to dhcp state machine?
[02:28:19] <Chris Donley> Advertise only?
[02:29:19] <Chris Donley> Can we say stop asking for IA_PD, but not IA_NA?
[02:29:23] <Chris Donley> JJB - yes
[02:30:00] <Chris Donley> JJB - need for incremental enablement
[02:31:10] <Chris Donley> Ralph - disagrees about granularity. Talk off-line.
[02:31:32] <Chris Donley> Ole - respin RFC 3315
[02:32:02] <Chris Donley> Ralph - don't see effect on state machine
[02:33:01] <Chris Donley> Ted - does not apply to information request? Mailing list
[02:33:36] <Chris Donley> 1) do we want to pursue solution 1, 2, or both? Top level or inside IA_NA, IA_PD, etc? Information requests?
[02:34:33] <Chris Donley> next preso - dhcp extension for port-set allocation
[02:34:38] kawashimam joins the room
[02:35:05] <Chris Donley> softwire lightweight 4over6
[02:35:53] <Chris Donley> next slide
[02:36:16] <Chris Donley> next slide
[02:36:29] kawashimam leaves the room
[02:36:42] kawashimam joins the room
[02:37:33] <Chris Donley> next slide
[02:38:05] <Chris Donley> need non-overlapping port set
[02:38:32] <Chris Donley> next slide
[02:39:34] <Chris Donley> next slide
[02:39:42] Suz leaves the room
[02:40:51] <Chris Donley> asking dhc wg for dhcpv4 port allocation option
[02:42:03] <Chris Donley> Ole - can be done via dhc and/or pcp. Reaching out to pcp, too?
[02:42:19] <Chris Donley> depends on operators
[02:42:57] <Chris Donley> Ralph - semantics - port sets on per-dhcp-client basis dhcp identifier to port set
[02:43:21] <Chris Donley> Ted - address+port set pairs
[02:43:31] <Chris Donley> 4rd?
[02:43:52] <Chris Donley> Make clear that it's not usable for host purposes
[02:44:13] <Chris Donley> scope down to nat usage
[02:44:54] <Chris Donley> Ted - not our problem
[02:46:46] <Chris Donley> yiu lee - architecture doc
[02:47:13] <Chris Donley> next - DHCP hint for PD
[02:47:57] <Chris Donley> overview
[02:48:29] kawashimam leaves the room
[02:48:29] kawashimam joins the room
[02:49:56] <Chris Donley> recommendation
[02:50:55] <Chris Donley> Hans - sounds like a bug in router. Don't think this is necessary.
[02:51:07] <Chris Donley> JJB - seeing this in the field.
[02:52:33] <Chris Donley> Ole - why would people with this bug support this, but not fix their broken implementations?
[02:53:31] <Chris Donley> need to document this problem
[02:54:44] <Chris Donley> Ralph - solution only works when CPE vendor recognizes bug, inserts hint, and doesn't fix bug
[02:55:34] <Chris Donley> JJB - Comcast sharing problems learned in the field.
[02:55:55] <Chris Donley> Ole - experiences with 6204 document?
[02:56:16] <Chris Donley> v6ops?
[02:56:49] <Chris Donley> Tim - hint not widely implemented in current implementations in iterops (5 of 17) null
[02:57:33] <Chris Donley> installing /60 on LAN instead of /64 - breaks LAN-side clients
[02:58:30] <Chris Donley> next preso
[02:58:39] <Chris Donley> dpchv6 multihoming
[02:59:12] <Chris Donley> inconsistent option format
[02:59:57] <Chris Donley> proposing exhaustive survey of dhcpv6 options and recommendations
[03:00:22] <Chris Donley> extend option guidelines draft? A little different
[03:01:40] <Chris Donley> show of hands: 8-9 people think this is worthwhile
[03:01:51] tsavo_work@jabber.org/Meebo joins the room
[03:02:03] <Chris Donley> next: DHCPv4 options still being added; option space getting crowded
[03:02:31] <Chris Donley> Option extension codes
[03:03:01] <Chris Donley> Ralph - how many are left? <40, maybe <20
[03:03:16] <Chris Donley> informally estimate burn rate
[03:03:53] <Chris Donley> Ralph has old draft
[03:05:35] <Chris Donley> looking for coauthor
[03:07:31] <Chris Donley> multihoming
[03:11:23] <Chris Donley> next slide
[03:12:09] <Chris Donley> 3315 doesn't consider multiple provisioning domains connected to same LAN (e.g. multiple ISPs)
[03:12:50] angga joins the room
[03:13:59] <angga> is there a link to the slides for this ? looked at http://datatracker.ietf.org/meeting/82/materials.html and it's not there
[03:15:05] <Chris Donley> Not posted yet. DHCPv6 and Multihoming Problem Statement, Ted Lemon, 10 minutes.
[03:15:25] <Chris Donley> Slides are photos of whiteboard
[03:16:06] <Chris Donley> flaw in DHCP
[03:16:09] <angga> yes....except that the pics are a bit hard to see (faint) from the back
[03:16:40] <Chris Donley> Ralph - when you say don't use DHCP in homenet, are you talking about prefix assignment?
[03:17:59] <Chris Donley> Ted - DHCP on both sides; need to fix this problem now before it gets worse.
[03:19:11] <Chris Donley> Why can't you tell that there are different prefixes? Node has to decide which prefix to prefer
[03:20:20] <Chris Donley> might need configuration from more than one server
[03:25:14] <Chris Donley> Ole - would this fundamentally affect architecture? client stops trying once it gets *one* response? Does this change it?
[03:26:17] <Chris Donley> First-hop might not reflect provisioning domains
[03:26:35] <Chris Donley> Ted - router needs to know it's talking to 2 provisioning domains and signal to node
[03:27:41] <Chris Donley> Ole: M flag?
[03:28:01] <Chris Donley> Ted - never figured out M/O bits because of this architectural flaw
[03:30:29] <Chris Donley> Jari - need to decide whether we want to solve it. Not convinced that homenet is a reason. Not a problem in Jari's house.
[03:32:35] <Chris Donley> Ole - some mechanisms e.g. RAs handle this situation (but not PD)
[03:33:23] fdupont leaves the room: Computer went to sleep
[03:34:48] <Chris Donley> Wes - take to mailing list
[03:34:59] angga leaves the room
[03:35:20] <Chris Donley> Ralph - clarifiy problem statement
[03:35:47] <Chris Donley> Wes - renumbering w/in enterprise space
[03:36:26] <Chris Donley> probably something that will come from 6renum
[03:36:59] Chris Donley leaves the room
[03:37:29] TACHIBANA toshio leaves the room
[03:37:37] kawashimam leaves the room
[03:38:59] kazubu leaves the room
[03:39:28] pselkirk leaves the room
[03:42:29] naito.kengo@jabber.org leaves the room
[03:50:59] tsavo_work@jabber.org/Meebo leaves the room
[03:53:37] Yuri Nawata leaves the room
[04:03:23] naito.kengo@jabber.org joins the room
[04:04:35] tatsuji leaves the room
[04:31:00] naito.kengo@jabber.org leaves the room
[04:31:22] Yuri Nawata joins the room
[04:44:34] TACHIBANA toshio joins the room
[04:48:38] angga joins the room
[04:48:50] angga leaves the room
[04:54:08] TACHIBANA toshio leaves the room
[05:00:41] kazubu joins the room
[05:04:08] Yuri Nawata leaves the room
[05:08:14] naito.kengo@jabber.org joins the room
[05:10:19] kazubu leaves the room
[06:23:16] naito.kengo@jabber.org leaves the room