Great outage of 1997 - Does anyone recall?
I recall a marvelous eighteen hour long global internet outage which I believe occurred in 1997, but this was before I'd ever touched BGP. Does anyone have the full story on this? I'm writing on article on the recent troubles with Supro and my silly editor wants fact checking and all sorts of stuff like that ... -- mailto:Neal@layer3arts.com // GoogleTalk: nrauhauser@gmail.com IM: nealrauhauser
On Feb 22, 2009, at 2:28 PM, neal rauhauser wrote:
Does anyone have the full story on this?
<http://www.merit.edu/mail.archives/nanog/1997-04/msg00444.html> ----------------------------------------------------------------------- Roland Dobbins <rdobbins@cisco.com> // +852.9133.2844 mobile Some things are just too precious to entrust to computers. -- Seth Hanford
On Feb 22, 2009, at 1:39 AM, Roland Dobbins wrote:
On Feb 22, 2009, at 2:28 PM, neal rauhauser wrote:
Does anyone have the full story on this?
<http://www.merit.edu/mail.archives/nanog/1997-04/msg00444.html>
Avi happened to be next to me when I read the first post in this thread - and re-read it out loud. I didn't even get to the end of the first sentence before he laughed and said "7007". (Avi & Vinny owned that together back then.) Operational content: Who still has 7007 filtered? -- TTFN, patrick
Does anyone have the full story on this? <http://www.merit.edu/mail.archives/nanog/1997-04/msg00444.html>
bottom line: o do not redistribute bgp into igp o do not redistribute dynamic igp into bgp o filter your peers and customers randy
On Feb 22, 2009, at 1:47 AM, Randy Bush wrote:
Does anyone have the full story on this? <http://www.merit.edu/mail.archives/nanog/1997-04/msg00444.html>
bottom line: o do not redistribute bgp into igp o do not redistribute dynamic igp into bgp o filter your peers and customers
And don't put all your most important infrastructure stuff (e.g. name server, mail server, shell host, etc.) in the first /24 of your / <shorter> allocation. The biggest problem with 7007 was not that it announced a bunch of prefixes. It is that 7007 announced _classful_ prefix (it had been filtered through RIP, remember?) with AS_PATH of ^7007$. This means if you had a 194.1.0.0/16, you saw 194.1.0.0/24 from 7007, which is more specific. Why this is bad is left as an exercise to the reader. And, of course, the problem persisted after the router in question was actually unplugged - not powered up or attached to any fibers/cables. Thank you Sprint for running beta code. :) -- TTFN, patrick
Well, I hope I'm not butchering the story up too badly - got an 800 word piece going up Monday on The Cutting Edge News and I'm doing something more lengthly and bloggy tonight for DailyKos, whilst hanging around abusing one of our spare 7507s with various new IOS versions. On Sun, Feb 22, 2009 at 12:55 AM, Patrick W. Gilmore <patrick@ianai.net>wrote:
On Feb 22, 2009, at 1:47 AM, Randy Bush wrote:
Does anyone have the full story on this?
<http://www.merit.edu/mail.archives/nanog/1997-04/msg00444.html>
bottom line: o do not redistribute bgp into igp o do not redistribute dynamic igp into bgp o filter your peers and customers
And don't put all your most important infrastructure stuff (e.g. name server, mail server, shell host, etc.) in the first /24 of your /<shorter> allocation.
The biggest problem with 7007 was not that it announced a bunch of prefixes. It is that 7007 announced _classful_ prefix (it had been filtered through RIP, remember?) with AS_PATH of ^7007$. This means if you had a 194.1.0.0/16, you saw 194.1.0.0/24 from 7007, which is more specific. Why this is bad is left as an exercise to the reader.
And, of course, the problem persisted after the router in question was actually unplugged - not powered up or attached to any fibers/cables. Thank you Sprint for running beta code. :)
-- TTFN, patrick
-- mailto:Neal@layer3arts.com // GoogleTalk: nrauhauser@gmail.com IM: nealrauhauser
On Feb 22, 2009, at 3:11 PM, neal rauhauser wrote:
Well, I hope I'm not butchering the story up too badly
This has been written up several times before - in addition to the links in Richard's post, take a look at the following, including the links at the bottom of the page: <http://lists.ucc.gu.uwa.edu.au/pipermail/lore/2006-August/000040.html> Here's a thorough writeup on the Supro incident: <http://asert.arbornetworks.com/2009/02/ahh-the-ease-of-introducing-global-ro...
For examples of specific applications of *deliberate* (as opposed to accidental, like AS7007) route hijacking, see the following: <https://www.defcon.org/images/defcon-16/dc16-presentations/defcon-16-pilosov...
<http://www.renesys.com/blog/2009/02/stealing-the-internet-back-1.shtml> <http://www.renesys.com/tech/presentations/pdf/blackhat-09.pdf> and then for extra credit, think about this: <http://www.blackhat.com/presentations/bh-dc-09/Marlinspike/BlackHat-DC-09-Ma...
<https://media.blackhat.com/bh-dc-09/video/Marlinspike/blackhat-dc-09-marlins...
and this: <http://www.unixwiz.net/techtips/iguide-kaminsky-dns-vuln.html> ----------------------------------------------------------------------- Roland Dobbins <rdobbins@cisco.com> // +852.9133.2844 mobile Some things are just too precious to entrust to computers. -- Seth Hanford
What was that story with an African routes some years back, any memories anyone? I am looking for a reference. On Sun, 22 Feb 2009, Patrick W. Gilmore wrote:
On Feb 22, 2009, at 1:47 AM, Randy Bush wrote:
Does anyone have the full story on this? <http://www.merit.edu/mail.archives/nanog/1997-04/msg00444.html>
bottom line: o do not redistribute bgp into igp o do not redistribute dynamic igp into bgp o filter your peers and customers
And don't put all your most important infrastructure stuff (e.g. name server, mail server, shell host, etc.) in the first /24 of your /<shorter> allocation.
The biggest problem with 7007 was not that it announced a bunch of prefixes. It is that 7007 announced _classful_ prefix (it had been filtered through RIP, remember?) with AS_PATH of ^7007$. This means if you had a 194.1.0.0/16, you saw 194.1.0.0/24 from 7007, which is more specific. Why this is bad is left as an exercise to the reader.
And, of course, the problem persisted after the router in question was actually unplugged - not powered up or attached to any fibers/cables. Thank you Sprint for running beta code. :)
-- TTFN, patrick
On Mon, Feb 23, 2009 at 12:06 AM, Paul Wall <pauldotwall@gmail.com> wrote:
On Sun, Feb 22, 2009 at 2:57 AM, Gadi Evron <ge@linuxbox.org> wrote:
What was that story with an African routes some years back, any memories anyone? I am looking for a reference.
146.20.0.0/16?
that's erie forge/steal... I think maybe Gadi's referring to the 41/8 used by an italian DSL provider for their internal network?? (not announced outside their ASN I don't think) -Chris
On Feb 22, 2009, at 10:10 PM, Christopher Morrow wrote:
On Mon, Feb 23, 2009 at 12:06 AM, Paul Wall <pauldotwall@gmail.com> wrote:
On Sun, Feb 22, 2009 at 2:57 AM, Gadi Evron <ge@linuxbox.org> wrote:
What was that story with an African routes some years back, any memories anyone? I am looking for a reference.
146.20.0.0/16?
that's erie forge/steal... I think maybe Gadi's referring to the 41/8 used by an italian DSL provider for their internal network?? (not announced outside their ASN I don't think)
Or the AFOL-KE thing with Above last March: <http://asert.arbornetworks.com/2008/03/africa-online-kenya-latest-internet-r...
-danny
On Sun, 22 Feb 2009, Danny McPherson wrote:
On Feb 22, 2009, at 10:10 PM, Christopher Morrow wrote:
On Mon, Feb 23, 2009 at 12:06 AM, Paul Wall <pauldotwall@gmail.com> wrote:
On Sun, Feb 22, 2009 at 2:57 AM, Gadi Evron <ge@linuxbox.org> wrote:
What was that story with an African routes some years back, any memories anyone? I am looking for a reference.
146.20.0.0/16?
that's erie forge/steal... I think maybe Gadi's referring to the 41/8 used by an italian DSL provider for their internal network?? (not announced outside their ASN I don't think)
Or the AFOL-KE thing with Above last March:
Thanks for all the references!
-danny
On Feb 21, 2009, at 10:28 PM, neal rauhauser wrote:
Does anyone have the full story on this?
See: http://www.flix.net/ -Richard
participants (9)
-
Christopher Morrow -
Danny McPherson -
Gadi Evron -
neal rauhauser -
Patrick W. Gilmore -
Paul Wall -
Randy Bush -
Richard Parker -
Roland Dobbins