20040126 21:37:59 | jabber@irc | jabber has joined the channel. 20040126 21:40:26 | jabber@irc | jabber has joined the channel. 20040126 21:54:55 | jabber@irc | jabber has joined the channel. 20040127 08:29:58 | gucci@irc | gucci has joined the channel. 20040127 08:31:01 | gucci@irc | gucci has left the channel. 20040127 13:29:35 | julien@irc | julien has joined the channel. 20040127 15:21:31 | julien@irc | julien quit IRC altogether 20040127 21:39:41 | jabber@irc | jabber has joined the channel. 20040127 21:44:43 | jabber@irc | jabber has joined the channel. 20040127 22:01:10 | koopal@irc | koopal has joined the channel. 20040127 22:09:38 | jabber@irc | jabber has joined the channel. 20040127 22:11:29 | jabber@irc | jabber has joined the channel. 20040127 22:14:36 | jabber@irc | jabber has joined the channel. 20040127 22:19:49 | jabber@irc | jabber has joined the channel. 20040128 00:16:26 | jabber@irc | jabber has joined the channel. 20040128 07:57:01 | bruce@jabber | *bruce* bruce has become available 20040128 08:03:52 | bruce@jabber | (laptop problems, looks like we're starting a little bit later) 20040128 08:05:46 | IanMarsh@irc | IanMarsh has joined the channel. 20040128 08:09:20 | emma@irc | emma has joined the channel. 20040128 08:10:03 | uk3@irc | uk3 has joined the channel. 20040128 08:11:13 | evilmark@irc | evilmark has joined the channel. 20040128 08:11:41 | evilmark@irc | morning 20040128 08:15:25 | iljitsch@jabber | *iljitsch* iljitsch has become available 20040128 08:16:26 | iljitsch@jabber | irc people, are you there? 20040128 08:16:40 | koopal@irc | no :-) 20040128 08:19:34 | evilmark@irc | yes 20040128 08:19:34 | shane_kerr@jabber | *shane_kerr* shane_kerr has become available 20040128 08:20:17 | iljitsch@jabber | Here is an explanation of the problem I just mentioned at the microphone: http://www.bgpexpert.com/archive2003q4.php#1 20040128 08:32:38 | marcoh@irc | marcoh has joined the channel. 20040128 08:32:46 | marcoh@irc | morning all 20040128 08:34:28 | bruce@jabber | 'sorry' ;) 20040128 08:35:35 | uk3@irc | marcoh: Sorry I did not make it yesterday, I was simply to tired 20040128 08:36:14 | hu60@jabber | *hu60* hu60 has become available 20040128 08:36:19 | marcoh@irc | uk3: nevermind 20040128 08:37:35 | koopal@irc | if nobody is relaying questions, I am willing 20040128 08:37:46 | koopal@irc | (providing there are people here not in the room) 20040128 08:38:33 | bruce@jabber | one of the things that struck me when implementing this is that people, who are in the room, would use it instead of getting up from their seats and walking to the mic. 20040128 08:41:30 | marcoh@irc | or starting a parallel discusion 20040128 08:42:58 | koopal@irc | that is a bigger risk 20040128 08:47:21 | marcoh@irc | see yesterdays anti-spam log 20040128 08:48:12 | mally@irc | mally has joined the channel. 20040128 08:48:17 | mally@irc | hi 20040128 08:48:26 | shane_kerr@jabber | Hi Mally. 20040128 08:49:00 | marcoh@irc | morning mally 20040128 08:50:56 | mally@irc | joao is v quiet on the mics... 20040128 08:51:17 | bruce@jabber | he isn't speaking into the mic. 20040128 08:51:23 | bruce@jabber | directly into. 20040128 08:53:35 | mally@irc | ah 20040128 08:54:00 | hu60@jabber | *hu60* hu60 has left 20040128 08:54:19 | shane_kerr@jabber | /shane_kerr test 20040128 08:57:02 | kurtis@irc | kurtis has joined the channel. 20040128 08:57:23 | koopal@irc | morning kurtis 20040128 08:57:52 | kurtis@irc | Morning! 20040128 09:07:54 | cron2@irc | cron2 has joined the channel. 20040128 09:09:06 | cron2@irc | so is this feedback thingie working? 20040128 09:09:25 | cron2@irc | HPH just reminded me that we forgot to announce it :-) (did *any* working group announce it?) 20040128 09:09:52 | koopal@irc | I haven't seen any workgroup announcing it myself 20040128 09:10:05 | emma@irc | it was in anti-spam 20040128 09:10:05 | mally@irc | anti spam did 20040128 09:10:11 | emma@irc | jinx ;-) 20040128 09:10:12 | evilmark@irc | aye 20040128 09:10:21 | shane_kerr@jabber | Not in routing or services. 20040128 09:10:27 | cron2@irc | at least *one* well-organized working group :-) *blush* 20040128 09:10:34 | kurtis@irc | Feedback? 20040128 09:10:46 | kurtis@irc | you mean the stuff on rosie? 20040128 09:10:57 | cron2@irc | kurtis: IRC channel that people could use to give us feedback from external hosts... 20040128 09:10:58 | mally@irc | irc.ripe.net/conference.ripe.net 20040128 09:11:08 | mally@irc | and irc.ripemtg.ripe.net (i think) 20040128 09:11:13 | kurtis@irc | Well, I forgott to hand out the whie-sheets o what would you expect? :-) 20040128 09:11:17 | bruce@jabber | http://rosie.ripe.net/network/feedback.html 20040128 09:11:28 | cron2@irc | mally: thanks, I found it (I'm here, after all :) ) 20040128 09:11:35 | kurtis@irc | cron2: I did announce that! 20040128 09:12:06 | koopal@irc | kurtis: then I must have been sleeping :-) 20040128 09:13:30 | mally@irc | i was going to say, i thought i remembered kurtis mentioning it... 20040128 09:14:19 | shane_kerr@jabber | apologies - I missed it too then. :( 20040128 09:20:22 | maribel@jabber | *maribel* maribel has become available 20040128 09:27:18 | maribel@jabber | *maribel* maribel has left 20040128 09:41:56 | jonl@irc | jonl has joined the channel. 20040128 09:45:59 | emma@irc | emma quit IRC altogether 20040128 10:03:08 | iljitsch@jabber | *iljitsch* iljitsch has left 20040128 10:03:35 | dede@irc | dede has joined the channel. 20040128 10:06:15 | jonl@irc | jonl quit IRC altogether 20040128 10:06:31 | jlawrence@irc | jlawrence has joined the channel. 20040128 10:07:50 | iljitsch@jabber | *iljitsch* iljitsch has become available 20040128 10:09:45 | ntitley@jabber | *ntitley* ntitley has become available 20040128 10:11:58 | evilmark@irc | coo more people 20040128 10:12:11 | cron2@irc | evilmark: is that a comment I need to relay to the microphone? 20040128 10:12:17 | evilmark@irc | don't think so, ta :) 20040128 10:12:54 | jlawrence@irc | how many of the people are actually logging in who aren't at the meeting :) ? 20040128 10:13:20 | ntitley@jabber | Not very many I suspect.... I'm certainloy at the meeting 20040128 10:13:26 | IanMarsh@irc | I'm not. 20040128 10:13:43 | ntitley@jabber | So that's 1 20040128 10:14:17 | jlawrence@irc | IanMarsh - are you picking up the audio feed as well, or just using the irc channel ? 20040128 10:14:52 | IanMarsh@irc | jlawrence: I'm attempting to view the video feed, but it's just started going all choppy on me. 20040128 10:15:17 | ntitley@jabber | Its not much better here... too much beer 20040128 10:29:53 | pim@irc | pim has joined the channel. 20040128 10:30:01 | dede@irc | dede has left the channel. 20040128 10:31:30 | evilmark@irc | do we need a document saying not to charge for IP addresses? 20040128 10:31:39 | pim@irc | pim quit IRC altogether 20040128 10:33:27 | IanMarsh@irc | Perhaps not a dedicated document, but it definitely should be a policy. 20040128 10:34:01 | jlawrence@irc | LIR's will simply not charge for IP addresses as such but will charge for the admin work that need to be done for getting the IP's. So essentially they won't charge for the IP's themselves but for the process of getting the IP's. 20040128 10:34:06 | iljitsch@jabber | arin does charge for v4 allocations at least, not sure about v6 20040128 10:34:22 | shane_kerr@jabber | ARIN only charges "end-users". 20040128 10:34:29 | shane_kerr@jabber | And then it's a fixed fee per allocation, not based on size. 20040128 10:34:30 | shane_kerr@jabber | IIRC. 20040128 10:34:33 | IanMarsh@irc | If an administrative charge is allowed, it should be capped, else LIRs can justify whatever cost they like as an administrative charge. 20040128 10:35:01 | iljitsch@jabber | the advantage of charging per address is that it levels the playing field. ripe membership is a barrier to entry for small isps. 20040128 10:35:12 | ntitley@jabber | It shouldn't be RIPE's responsibility to set maximum fees for administration, IMHO 20040128 10:36:16 | cron2@irc | I agree with nigel 20040128 10:36:54 | IanMarsh@irc | That's fine so long as there's plenty of competition; then consumers will pick the provider who offers IPs with their connection for a lower fee. 20040128 10:37:13 | iljitsch@jabber | shane: see ISP Subscriptions for Bulk IP Registrations at http://www.arin.net/registration/fee_schedule.html 20040128 10:37:55 | jlawrence@irc | Admin charges are and need to vary greatly - ie there's a good chance that in one country admin staff cost an LIR more than in another country. 20040128 10:38:22 | ntitley@jabber | Agrred 20040128 10:39:46 | jlawrence@irc | I think that charging should be left to market forces to decide the amount to charge. But it needs to be specified that you're only RENTING the IP space not buying them. 20040128 10:39:56 | IanMarsh@irc | Agreed. 20040128 10:40:16 | IanMarsh@irc | So long as market forces are strong enough to keep it low. 20040128 10:40:38 | ntitley@jabber | Yes, as has been raised here, otherwise folk start to think they own address space and want to take it with them, like domain names 20040128 10:41:30 | iljitsch@jabber | haven't we learned nothing from the telcos? they still charge by the minute in most places and billing costs more than the service being billed. high risk of the same for ip addresses 20040128 10:41:46 | IanMarsh@irc | Didn't someone mention that in Africa, they're short of LIRs, and thus those LIRs charge a very large fee. 20040128 10:42:08 | koopal@irc | do you want that comment relayed in the meeting? 20040128 10:42:35 | IanMarsh@irc | I think charge-per-minute isn't a valid comparison. 20040128 10:42:50 | jlawrence@irc | There are enough LIR's around the RIPE region to keep things low and I think that it'll go the same as name space - some registrar's are dirt cheap others are expensive. 20040128 10:43:56 | jlawrence@irc | If in Africa ISP's are only getting a /27 then they are bound to want/need to charge for IP's as for them the IP's are actually scarce. 20040128 10:44:53 | iljitsch@jabber | just recently I commented on the IETF list that broadband provider putting their customers behind NAT isn't done in more civilized parts of the world (= in europe) but I got two messages from europeans that saw this happen. The trouble is that most people aren't informed consumers and they're ruining it for those of us who are. 20040128 10:45:37 | IanMarsh@irc | Standard supply vs demand, then. If we increase the supply, the 'price' should drop? 20040128 10:46:11 | iljitsch@jabber | there are 1400000000 addresses laying around in the IANA warehouse, with about 60000000 being given out per year = not scarce so any scarcity has little to do with addresses per se 20040128 10:46:19 | cron2@irc | it's an education thing as well. some african countries just *assume* that addresses are scarce, without even talking to the registries 20040128 10:46:29 | cron2@irc | this partly political nonsense 20040128 10:46:39 | evilmark@irc | they are made artificially scarce by expecting people to justify usage 20040128 10:46:51 | iljitsch@jabber | and the media keep sending out the wrong message. see apnic they went throughh the trouble of correcting a bbc story 20040128 10:47:14 | cron2@irc | justifying usage doesn't sound wrong to me, if there is no more price tag than "fill in this form" 20040128 10:47:20 | ntitley@jabber | That was actually a joint correction by the LIRS together 20040128 10:47:35 | ntitley@jabber | Justifying usage isn't too difficult 20040128 10:47:39 | cron2@irc | there seems to be consensus to make the document "historic" 20040128 10:47:40 | iljitsch@jabber | maybe we should just give them to anyone who asks so they're gone in 3 years and we can finally start using ipv6. :-) (half kidding) 20040128 10:47:48 | jlawrence@irc | It's not just the Africans that assume things are scarce - the general consumers also do (if they even know what an IP address is). 20040128 10:47:52 | koopal@irc | cron: isn't that lack of training towards the LIR's as well? 20040128 10:47:53 | cron2@irc | iljitsch: *g* 20040128 10:48:08 | dfk@jabber | *dfk* dfk has become available 20040128 10:48:32 | cron2@irc | koopal: the LIR I've personally experienced this with (tunesia) was very stubbornly refusing to be trained. Timothy Lowe from the NCC was *there* and very clearly explained *there is no shortage* - but they just didn't believe him 20040128 10:48:44 | koopal@irc | hmmm 20040128 10:48:50 | ntitley@jabber | Possibly they don't *want* to believe it 20040128 10:48:55 | cron2@irc | political/cultural thingie 20040128 10:48:58 | cron2@irc | power play 20040128 10:49:10 | cron2@irc | (all ISPs in tunesia need to go through the one central LIR by law) 20040128 10:49:13 | ntitley@jabber | Its quite nice to be in the position of being perceived to be the source of a scarce resource 20040128 10:49:28 | koopal@irc | interesting then that afrinic starts with a /22 allocation instead of /21 20040128 10:49:36 | ntitley@jabber | Whish is what cron2 is saying 20040128 10:49:43 | cron2@irc | koopal: yes... 20040128 10:49:58 | dfk@jabber | to nigel: So now I understand why we all insist we have "clue"! ;-) 20040128 10:50:09 | ntitley@jabber | :-) 20040128 10:50:33 | ntitley@jabber | Clue is a similarly scarce resource 20040128 10:50:47 | cron2@irc | ntitley: definitely :( 20040128 10:51:12 | iljitsch@jabber | hmmm store room next to iana belongs to the rfc editor... plentyy of clue in stock 20040128 10:51:22 | jlawrence@irc | But we're supposed to be the ones with a clue - sometimes I wonder :) 20040128 10:52:34 | cron2@irc | icann is also quite resistant to picking up clue 20040128 10:52:48 | shane_kerr@jabber | Right now we hand out IP addresses based on "need". Is setting up a business to sell IP addresses "need"? 20040128 10:52:52 | emma@irc | emma has joined the channel. 20040128 10:53:23 | jlawrence@irc | Is there any consensus at what point the IPv6 gains a critical mass and starts to take over. 20040128 10:53:37 | ntitley@jabber | Shane: We actually hand out IP addresses based on ability to fabricate RIPE forms 20040128 10:53:41 | dfk@jabber | icann is improving, hey they hired crain ;-) and twomey ;-), but it may not be fast enough to save them 20040128 10:53:43 | cron2@irc | jlawrence: it might or might not happen... 20040128 10:53:54 | IanMarsh@irc | ntitley: How true, how true. 20040128 10:54:02 | cron2@irc | I'd say "if a standard end user will do their daily surfing mostly via IPv6 without noticing", the critical point is reaching 20040128 10:54:10 | cron2@irc | dfk: they are still giving /23s to RIPE 20040128 10:54:15 | jlawrence@irc | Ntitley - so I'm not the only one who thinks everyone lies to get their allocation/assignments 20040128 10:54:29 | shane_kerr@jabber | /me scribbles down some names. 20040128 10:54:30 | shane_kerr@jabber | ;) 20040128 10:54:33 | ntitley@jabber | jlawrence: most do 20040128 10:54:42 | dfk@jabber | from my experience as a hostmaster: lying consistently is much harder than one expects 20040128 10:55:05 | ntitley@jabber | Depends on how long you've been doing it, and how good your records are :-) 20040128 10:55:17 | dfk@jabber | about adress in exchange for filled in forms: suggest a better system! 20040128 10:55:18 | ntitley@jabber | And I'm an ex-hostmaster too 20040128 10:55:30 | koopal@irc | dfk: that is valid for large assignments I think 20040128 10:55:31 | cron2@irc | dfk: yep, usually the numbers are just "too right" if you're laying 20040128 10:55:34 | jlawrence@irc | dfk - I don't think so. So long as you keep records and are consistent. 20040128 10:55:35 | koopal@irc | but not for small 20040128 10:55:36 | dfk@jabber | ntitley: but you have "clue" 20040128 10:55:51 | ntitley@jabber | Ahh, so now we trade addresses for clue 20040128 10:56:00 | cron2@irc | that's a fine currency to me 20040128 10:56:32 | ntitley@jabber | Good, so now we just have to test for clue, and hand out addresses appropriately 20040128 10:56:41 | dfk@jabber | for conversation one does not care about the small fry. I am all for /21 in exchange for RIR fee 20040128 10:56:43 | ntitley@jabber | I think we are back to RIPE forms 20040128 10:57:02 | ntitley@jabber | dfk: yes I agree 20040128 10:57:04 | cron2@irc | dfk: that's what we have now 20040128 10:57:18 | dfk@jabber | We really need a global discussion COVERSATION vs AGGREGATION taking into account current "reality" 20040128 10:57:19 | ntitley@jabber | And long overdue 20040128 10:57:28 | cron2@irc | dfk: yes 20040128 10:57:30 | shane_kerr@jabber | As SW department manager, I'm hoping the LIR Portal will reduce the required clue to complete an IP allocation request. 20040128 10:57:35 | ntitley@jabber | Yes, I agree with your comment this is long overdue 20040128 10:57:37 | iljitsch@jabber | jlaw: what is the target we want to meet? I don't think we're going to see more v6 than for v4 for general web browsing and other common stuff any time soon, but as soon as 10% of all people use ipv6 for 10% of their traffic I think we can conclude that IPv6 meets a need 20040128 10:58:00 | iljitsch@jabber | dfk: what's your take on this reality? fortunately the global routing table isn't growing all that fast anymore 20040128 10:58:20 | ntitley@jabber | And the routers can handle it these days anyway 20040128 10:58:22 | jlawrence@irc | If you're being truthful in your request, I think you need little 'clue' to fill in the allocation request. 20040128 10:58:35 | dfk@jabber | my take is that we should do less conservation and more aggregation 20040128 10:58:46 | cron2@irc | jlawrence: have you seen applications by clueless customers recently? 20040128 10:59:09 | ntitley@jabber | dfk: I agree 20040128 10:59:10 | cron2@irc | dfk: that was one of my motivations behind the last change: go away from startup-LIRs having to request multiple PIs due to not being able to get a PA block 20040128 10:59:32 | jlawrence@irc | I have many clueless customers - I'm still arguing with some that they need/must start using http1.1 20040128 10:59:33 | iljitsch@jabber | yes and people can survive jumping out a window 1 storey high but that isn't a reason to start doing it. bgp convergence is far from ideal but there is no impending train wreck just yet 20040128 10:59:46 | dfk@jabber | my take on the foreseeable future in layer3&4: v4 addresses in layer 3 and v6 addresses in layer 4 20040128 11:00:09 | shane_kerr@jabber | dfk: ??? 20040128 11:00:27 | cron2@irc | 6to4? 20040128 11:00:35 | ntitley@jabber | Dfk is right, he's talking about v6 being tunneled under v4 20040128 11:00:39 | iljitsch@jabber | dfk: i think a good start would be to aggregate where possible, which still isn't done to the degree desirable, and downward pressure on allocation sizes is very bad 20040128 11:01:21 | jlawrence@irc | How many of the broadabnd routers actually support v6 natively ? 20040128 11:01:22 | iljitsch@jabber | if /22s are common in africa that pretty much means that everybody gets to use them too because filtering on the specific allocation size for each block is quickly becoming a full time job 20040128 11:01:32 | dfk@jabber | not quite: i am talking about v6 addresses as (transport) end point IDs and v4 addresses for routing packets .... 20040128 11:01:39 | ntitley@jabber | OK 20040128 11:01:52 | iljitsch@jabber | what's a broadband router? 20040128 11:01:56 | dfk@jabber | ... wherever routing matters, e.g. to my DSL modem etc. pp. 20040128 11:02:19 | cron2@irc | dfk: some DSL provider in .NL and .DE have started offering native IPv6 in parallel to native IPv4 20040128 11:02:32 | dfk@jabber | ... it does not involve tunnels. 20040128 11:02:51 | ntitley@jabber | Yes, I understand 20040128 11:03:05 | jlawrence@irc | iljitsch - I suppose it's a bradband modem with more facilities. 20040128 11:03:06 | dfk@jabber | cron2: I use that service too. I just do not think it is logical to do. 20040128 11:03:13 | cron2@irc | why? 20040128 11:03:55 | dfk@jabber | cron2: no added features for the customer .... and no added features for the ISP either (routing is the same) 20040128 11:05:01 | jlawrence@irc | AFAIK there are only a couple of ISP's offering native v6 in the UK. 20040128 11:05:33 | dfk@jabber | jlawrence: yes and it is going to fly when microsoft turns it on by default ;-) 20040128 11:05:53 | jlawrence@irc | Or will it crash like most of their other stuff :) 20040128 11:06:12 | cron2@irc | dfk: are you talking a specific protocol proposal (6to4, teredo) or something new? 20040128 11:07:11 | dfk@jabber | cron2: just a general feeling about the architecture and the things that drive it from 300.000 feet up 20040128 11:08:52 | bruce-hp@jabber | *bruce-hp* bruce-hp has become available 20040128 11:09:20 | iljitsch@jabber | jlaw: i have a cisco 827 (of 826, i forget) at home that does v6. smaller aggregation boxes such as cisco 7200 have no trouble with it either but many others don't. 20040128 11:09:53 | iljitsch@jabber | jlaw: Apple has v6 on by default. But their browser won't use it... 20040128 11:10:25 | bruce-hp@jabber | won't or isn't configured by defaut? 20040128 11:10:36 | shane_kerr@jabber | Won't now. 20040128 11:10:37 | dfk@jabber | I ose osx and I turn it off because I feel uncomfortable about it security wise, firewall etc ... 20040128 11:10:45 | shane_kerr@jabber | Next version should give users an option. 20040128 11:13:06 | jlawrence@irc | Few cpe's are cisco boxes with home users (certainly in the UK). I use an Intertex IX66 which works well but means that I'll have to tunnel v6. Ordinary customers will not be able to setup tunnels - they have enough problems setting up email clients ;) 20040128 11:13:19 | iljitsch@jabber | dfk: they have a v6 firewall also now but don't worry as apple doesn't expose services to the network unless you turn it on and none of those except afs do v6 anyway 20040128 11:13:50 | iljitsch@jabber | jlaw: if you don't do nat 6to4 is a breeze 20040128 11:13:53 | cron2@irc | as for the CPE: yes, that's a problem. We try to roll out more "bridged ethernet" CPEs, so the CPE doesn't actually need to knwo any L3 protocols 20040128 11:13:58 | ntitley@jabber | jlaw: as a matter of interest which UK providers do IPv6 native at the moment? 20040128 11:14:29 | jlawrence@irc | Andrews and Arnold that I know of - I think you have to request it though. 20040128 11:15:15 | ntitley@jabber | jlaw: thanks 20040128 11:16:12 | jlawrence@irc | iljitsch - try explaining to a customer what 6to4 is :). 20040128 11:16:52 | jlawrence@irc | cron2 - another problem in the UK is that we use PPPoA which makes bridging more difficult. 20040128 11:16:55 | cron2@irc | iljitsch: just read the -261 document. WIth sparse allocation, you have no buffer bit problem anymore 20040128 11:17:43 | cron2@irc | jlawrence: we have PPPoE or sometimes bridged 1493 ATM - the latter is easy, PPPoE won't work with bridging either (but then you can get DSL modems that offload PPPoE to a linux client) 20040128 11:17:47 | cron2@irc | yes, it's messy 20040128 11:18:00 | cron2@irc | ... which is why our service doesn't really take off (some 20 customers by now) 20040128 11:19:40 | jlawrence@irc | cron2 - I thought PPPoE could be bridged with out to many problems. 20040128 11:22:29 | cron2@irc | jlawrence: "sort of" - you bridge the PPP packets, but it means that you need to have a client PC (or so) that can do PPPoE with IPv6CP. The DSL modem is passive in this case. 20040128 11:22:42 | cron2@irc | but as WinXP PPPoE doesn't do IPv6 (but they can do it via LAN) this isn't helpful... 20040128 11:30:09 | jabber@irc | jabber has joined the channel. 20040128 11:30:36 | bruce@jabber | *bruce* bruce has become available 20040128 11:31:02 | bruce@jabber | sorry about that; the jabberd process ran out of memory :( 20040128 11:31:23 | bruce-hp@jabber | *bruce-hp* bruce-hp has become available 20040128 11:31:53 | cron2@irc | could someone else please tell joao that the management address does NOT need to be a spearate /24 that's globally visible? 20040128 11:32:54 | jlawrence@irc | Nearly time for lunch :) 20040128 11:33:17 | cron2@irc | iljitsch: IPv6 is different - there are no well-established filtering rules yet, but for IPv4, we have... 20040128 11:33:19 | iljitsch@jabber | *iljitsch* iljitsch has become available 20040128 11:33:30 | cron2@irc | oops, need to repeat that 20040128 11:33:31 | cron2@irc | iljitsch: IPv6 is different - there are no well-established filtering rules yet, but for IPv4, we have... 20040128 11:34:38 | iljitsch@jabber | quite the opposite. in v6 the rule is extremely simple: filter on /32. in v4 it's a big mess so basically "anything below /24 is dead in the water and the published rir size is always ok" but there is WAY too much room for nonsense in between 20040128 11:34:57 | jlawrence@irc | Unless someone starts building routers with shed loads of memory, we could easily end up with massive routing tables. 20040128 11:35:27 | jlawrence@irc | if we filter on a /32 how big is the routing table going to get ? 20040128 11:35:35 | IanMarsh@irc | Naive suggestion: can a /xx be put aside from which /32 assignments for anycast usage can be made, and since these will be used for major services such as DNS, surely providers can then exempt the whole /xx from their filters? 20040128 11:35:41 | dfk@jabber | *dfk* dfk has become available 20040128 11:35:45 | shane_kerr@jabber | jlawrence: 2^32 20040128 11:35:46 | shane_kerr@jabber | ;) 20040128 11:35:47 | iljitsch@jabber | see http://www.ripe.net/ripe/docs/smallest-alloc-sizes.html 20040128 11:36:10 | jlawrence@irc | exactly 2^32 is still too big to fit in my routers :) 20040128 11:36:55 | iljitsch@jabber | this seems very nice until you realize that RIPE actually gives out blocks the size of the minimum ASSIGNMENT so for 3 /8s you need to accept /27s = 3 x 2^19 = 1.5 million potential routes 20040128 11:37:09 | jlawrence@irc | if arin are giving /48 's to root servers, are we going to have a single resource where we can find out exactly what /48's we will need to allow through our filters ? 20040128 11:37:28 | shane_kerr@jabber | I think they're all from a single /32. 20040128 11:37:41 | iljitsch@jabber | 2^32 = 4G at about 60 bytes per route (milage may vary) = a 24 GB BGP table. FIB is even bigger 20040128 11:38:48 | IanMarsh@irc | 4 x 60 = 240 20040128 11:39:27 | jlawrence@irc | Anyone got a router that can handle that - and that's only assuming 1 set of BGP tables. 20040128 11:39:27 | iljitsch@jabber | also not very pleasant: routing table processing scales at the order of O(n * log n) which is not good 20040128 11:39:37 | iljitsch@jabber | lanmarsh: sorry 20040128 11:40:27 | iljitsch@jabber | jlaw: http://www.arin.net/policy/2001_3.html (this is pretty hard to find) 20040128 11:41:58 | IanMarsh@irc | :) Just I want a 240Gb routing table considerably less than I want a 24Gb one! 20040128 11:42:18 | iljitsch@jabber | wait this isn't the actual list... 20040128 11:42:18 | iljitsch@jabber | http://www.arin.net/registration/ipv6/micro_alloc.html 20040128 11:43:08 | cron2@irc | as for the routing table: I suggested to give a "one size fits all" /20 to everbody (capping the total number of routes to 2^17 = 128.000), which would solve *that* - but people are more keen on conservation -> /32 allocations 20040128 11:43:18 | jlawrence@irc | exellent thanks iljitsch 20040128 11:44:14 | jlawrence@irc | can't wait to get my hands on a router which can handle 240Gb tables - don't think I'd like to see the price :) 20040128 11:45:03 | shane_kerr@jabber | /me imagines a G5 using a 300 GB disk as swap space, running Zebra... 20040128 11:45:27 | iljitsch@jabber | shane: I've been thinking about an altivec optimized patricia tree implementation. :-) 20040128 11:45:35 | jlawrence@irc | holding that tables in a swap would make the router too slow :) 20040128 11:46:15 | jlawrence@irc | It might be an excuse to start playing with Opteron's 20040128 11:46:21 | cron2@irc | back to route cache 20040128 11:46:27 | iljitsch@jabber | jlaw: don't even kid about that. old zebra issue: kernel routing table won't fit in the 20 MB space allocated in old freebsd kernels 20040128 11:46:55 | bruce-hp@jabber | depends whether your focus is packets or streams. If you cache the current streams, you'd have a huge overhead for an initial stream, but then it hits your stream cache 20040128 11:47:05 | bruce-hp@jabber | (and this has already been done) 20040128 11:47:39 | iljitsch@jabber | bruce: ever heard of sql slammer & friends? 20040128 11:47:46 | cron2@irc | bruce-hp: and the next DoS/Portscan/SQL slammer will kill you 20040128 11:47:50 | cron2@irc | iljitsch: exactly 20040128 11:48:31 | bruce-hp@jabber | cron2; yup. 20040128 11:48:46 | bruce-hp@jabber | hence, you don't do it that way. 20040128 11:49:08 | bruce-hp@jabber | 'do you want lunch or do you want ipv6?' ;) 20040128 11:49:30 | jlawrence@irc | It looks like lunch is the answer :) 20040128 11:49:55 | shane_kerr@jabber | *shane_kerr* shane_kerr has left 20040128 11:50:29 | iljitsch@jabber | see you at lunch! 20040128 11:50:47 | jlawrence@irc | jlawrence quit IRC altogether 20040128 11:50:57 | bruce-hp@jabber | *bruce-hp* bruce-hp has left 20040128 11:51:07 | iljitsch@jabber | *iljitsch* iljitsch has left 20040128 11:51:27 | IanMarsh@irc | IanMarsh quit IRC altogether 20040128 11:52:22 | uk3@irc | uk3 quit IRC altogether 20040128 11:52:37 | emma@irc | emma quit IRC altogether 20040128 12:00:35 | dfk@jabber | *dfk* dfk has left 20040128 12:06:38 | dfk@jabber | *dfk* dfk has become available 20040128 12:13:04 | dfk@jabber | *dfk* dfk has left 20040128 12:39:47 | uk3@irc | uk3 has joined the channel. 20040128 12:58:38 | evilmark@irc | evilmark has left the channel. 20040128 12:58:50 | koopal@irc | koopal has left the channel. 20040128 13:00:44 | gucci@irc | gucci has joined the channel. 20040128 13:02:39 | marcoh@irc | marcoh has left the channel. 20040128 13:06:07 | uk3@irc | uk3 quit IRC altogether 20040128 13:09:47 | cron2@irc | cron2 has left the channel. 20040128 13:52:37 | gucci@irc | gucci quit IRC altogether 20040128 13:53:38 | mally@irc | mally has left the channel.