| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
Comments:
<0> remember about 5 years ago when an ISP accidentally polluted the core routing tables, making their T1 look like the best link on the net?
<1> Strider: I think that sales dude has been talking to my boss ;)
<2> :)
<0> it was sadly amusing, and demonstrated how fragile the 'net really is.
<2> :)
<0> it took down most of the internet in north america, just by polluting the core routing tables.
<1> doesn't surprise me. BGP has built-in support for failing to believe untrustworthy routers. unfortunately, it can't protect against trustworthy routers that have been fscked up by their operators
<3> Strider, wasn't it longer than five years ago, and wasn't it NAPnet in Chicago?
<0> I don't know...
<3> Strider, or are we remembering different routing table ****age events? ;)
<0> IceKarma: could be.. I just have this vague recollection of an ISP in the US nuking traffic for most of the 'net.
<0> heh, when i was working on the terabit fabric we sort of sat there and went.. "How the **** do you bring in a terabit of data into one rack anyhow?"
<2> DaveHowe i think that cisco has a solution for the problem of BGP no?
<0> we can build a fabric that fast...
<3> Strider, lots and lots of fibre?
<0> arpf: sure they do...
<0> if you've got the bank.
<1> arpf: well, their sales droids would tell you they do, but look shifty if you press them for details
<1> normally, they just sell you bigger ciscos with more ram in them :)
<0> DaveHowe: heh.. i've been known to make, on more than one occasion, the finger-slip of "crisco"
<2> well may be but i have an article here discrabing in details the solution
<3> Strider, ever done it to their face, though?
<0> arpf: articles are a dime a dozen.. any good marketing droid can write up an article in 30 minutes.
<3> Strider, I was on the phone with the Cable & Wireless NOC once and called them "Cable and Witless" to an engineer ;)
<0> IceKarma: a few times. :P
<2> i can give you the link guys and you can analyse it
<1> Strider: someone jokingly told a cisco rep that their memory was too expensive and he was going to "rip some out of his old commodore 64 and put that in instead"
<0> DaveHowe: haha
<4> I only analyze .. Z == cooler.. and correct.
<3> DaveHowe, Sun NZ wanted NZD9k for 128 MB of RAM for an SS20 in 1996
<3> DaveHowe, Kingston wanted NZD4500 for the same
<1> Strider: next thing he knew, cisco canceled his support contract and said it could only be reinstated after he paid the full fee *again* and a cisco engineer came out to verify no "non authorised memory" was in the supported device
<0> IceKarma: apple used to be the same way.
<1> Strider: rant was on ASR about it, iirc
<3> DaveHowe, sheesh.
<0> DaveHowe: mmm.. scary devil monastary
<4> how else is cisco going to guarantee that the government can tap in and do what they want with the internet ?
<1> IceKarma: doesn't surprise me. SUN memory was pretty bad, and HPUX memory was even worse
<2> shame on them guys at cisco!!
<3> does anyone remember the Cisco support engineer who came in here the night he knocked a Cisco 12000 out a fourth-storey window? ;)
<0> DaveHowe: i'll bet you had to change the license for HPUX if you wanted it to address more RAM
<1> Strider: nope. but oracle would stop working until you bought a new licence for it
<0> ahh
<0> yeah
<1> arpf: cisco like big bgp tables - it means they can sell bigger ciscos to handle it, and the invariable space needed when the routes flap after an isp goes under...
<2> i'm asking now guys at bgpexpert and they fuc** told me that the damn BGP problem is resolved what's the sh*t is going on here????
<1> which "bgp problem" do they mean though?
<1> the "big" bgp problem in their eyes was that we were about to exhaust the AS address space
<2> ipv6 of course and the damn cisco solution
<1> ( I say "about" - we had about a third of them left)
<0> i thought MPLS was cisco's solution.. or was that last week?
<2> yeas Strider
<2> TES
<2> YES
<4> easy fix for the next decade for ipv4, get rid of the A block holders that dont require the A block. like MIT, us military, IBM etc.
<2> its mpls
<5> Strider: Come to FOSDEM
<5> safemode, DaveHowe: You too
<0> FOSDEM?
<5> http://www.fosdem.org/
<0> rhowe: wrong continent.
<5> pfft
<5> Hardly matters these days
<5> Check flight prices
<0> rhowe: heh
<0> plus there's better things to do with my vacation time then hang out with a bunch of dorks. :P
<1> safemode: yeah, but all the firms that were willing to hand back Cl*** As suddenly found the cl*** Bs or worse they got back had to be paid for and weren't as routable
<1> safemode: leading to an understandable reluctance on the part of the others
<0> iirc, the us.mil owns like 60 cl*** As
<4> aww... you gotta pay for it? those poor poor companies
<4> my heart bleeds for them
<1> safemode: sure - but if you are handing back a cl*** A you don't have to pay for, and suddenly find yourself billed for the favour (plus the expense of having to renumber your space)....
<2> DaveHowe even dhcp and nat can't resolve the problem of the limitation of ip cl***es?
<2> don't forget NGNs
<1> dhcp isn't an answer, its a fudge. NAT is ok, but you can't easily run servers behind NAT
<2> that's the main problem DaveHowe
<2> nat ****s
<0> nat is a hack..
<0> i don't ocnsider dhcp to be a hack, more of a management tool
<2> so the solution is ...
<0> dynamic address ***ignment has been around since the sawn of time.
<2> ipv6
<5> DaveHowe: Excuses, excuses :)
<0> arpf: yes, but for it to be adopted there must be an economic driver for it in the next quarter.
<0> if not, it will never happen.
<1> Strider: its a fudge. it means you can ***ign a limited address space more efficiently by ***uming only a fraction of the nodes will need addresses at any time, and will hand them back cleanly enough they can be re-used
<0> DaveHowe: that is one of its capabilities.. that's not why it was developed
<1> Strider: but the downside is that you can't guarantee a node won't change ip address, and dns is really too slow to keep up
<0> it was developed to alievieate subnet portability in routed networks and also to alieviate the administrivia of having to ***ign things manually to devices that are moved around.
<0> DaveHowe: sure you can.
<0> DaveHowe: by either a) playing with the lease times, or b) hard-***igning an IP to the host.
<0> all my core machines are dhcp ***igned except for the DCHp server itself, but they're all hard-wired to their IP
<1> Strider: you can do that for a limited number of nodes, but if your node count exceeds your address space, you can't possibly do it for all of them
<0> moves the administration to the server, which is easier to handle than having to go and change all the hosts on the network.
<0> DaveHowe: right, but that isn't why DHCP and bootp for that matter was developed.
<1> Strider: yeah, that's a benefit - but the downside there is that it doesn't help towards ip address space exhaustion, which was the point made originally
<0> they were developed to simplify and centralize administration.
<0> and reduce the support load.
<5> IP addresses aren't hard to get IME
<1> well, bootp was more to cheapen machine manufacture by omitting bits of it :)
<5> I have a /29 at home, and at work we have at least a /27 and a /28
<4> 5 bucks a pop
<4> usually.
<5> Should be getting a /24 too, all for 'free' with our ISP
<1> although I use PXE to push out installations with Zen
<5> Still, they are our ISP's addresses, not outs
<3> room for future expansion ;)
<0> heh, I have a /32 ;)
<0> but meh
<3> I used to have a /24
<4> in the end, you still dont get for peer access to what you pay for, unless you pay a whole lot. So basically the internet is casterated no mattter what the state of IP's is for the majority of people. IT's really sad.
<3> now a friend running an ISP in NZ is using it
<0> safemode: though sadly, that's a feature now.. most people are incapable of maintaining secure hosts so the lack of peer access is a benifit.
<2> and if we distrebute ip adresses for 5 buks each i think that we'll have a huge problems
<2> and if we plan to connect all over ip in the near future we must find a solution, and i think that we will not have a choice to use ipv6
<0> arpf: right, but customers don't count.
<0> it won't be long before ISPs art using NAT themselves..
<0> hell, I think mos tpeople are lucky that they can still get a publically accessable ip.
<0> is it the right hting to do? no, absolutely not.. but they'll do it because it's cheaper than going to ipv6t
<0> and that's all that matters.
<1> plus of course, most of those in power want a consumer internet - TV model, big companies run servers, normal users pay for access to copyrighted media
<1> NAT to the edge nodes helps that, by preventing those nodes being p2p or servers, just strict consumers
<0> exactly
<0> in th eend the consumer gets screwed but that's the way it goes.
<2> yes
<1> the consumer never matters - as long as he stays a consumer. the idea that customers could be *competitors* scares most big companies
<1> its bad enough they are *allowed* to post information about the big companies on websites they can't control and censor
<1> and if musicians ever find out they no longer need record companies to distribute music for them...
<2> but the thing that i couldn't understand ,why universities and research centers are investind millions in ipv6 projects???
<1> they aren't. however, Internet 2 was a big thing in american unis up to about a year ago.seems to have gone quiet on that though
<0> arpf: because it's cool(tm)
<2> so they don't care now to find a better solution then ipv6???
<0> arpf: because they realize it's futile, probably.
<2> :)
<1> arpf: no. some just want to find the Next Big Thing so they can patent it and make a fortune on royalties
<1> most just because research is what universities DO - they take money and generate papers :)
<3> I'm already half-screwed, the only P2P system that seems to work over my ISP is BitTorrent
<2> you're right DaveHowe
<1> many block it now. emule has an encrypted mode to work around it, with distributed dbs, but its not that reliable
<6> hi
<6> i have bit of trouble enabling my SBC realtek8185L WiFi, somebody experienced ?
<2> morpheus and limewire are working fine
<7> I'm in lamaz cl*** right now.
<7> I'd rather be in #linux :/
<1> limewire has too many false positives and deliberate plants - usually trojans or riaa bombs
Return to
#linux
or
Go to some related
logs:
uninstall plesk suse
cannot find symbol method getdocumentbase
compile samba ldapcontrol
#chatzone
ubuntu apache change error_log
george carlin hippy dippy weather man
EAGLE Chili Sauce
#mirc
lofl
wolfrage undernet
|
|