Re:BGP and RPKI
by tialaramex on Saturday April 18, 2020 @02:48PM
(#22910099)
I have been an A&A customer for years and so this is disappointing because it's exactly the sort of crappy response I'd criticise other ISPs for. Hopefully in the not too distant future we can talk some sense into RevK and co.
The choice to block one route to make isbgpsafeyet apparently pass is bogus, it's like silencing a fire alarm (which is also something I see far too many businesses choosing to do, oh Zone A keeps alarming, we'll just disable it even though now that means more danger, engineer calls are too expensive)
Those familiar with my work will know that I read a lot of accident investigation reports, and a recurring theme in maritime accidents involving large vessels is that the BNWAS (a technology which alerts sleeping crew elsewhere on a vessel that for some reason nobody on the bridge is paying attention, probably because they're asleep) is "somehow" muted, with its logging disabled or even entirely switched off during an incident. That might not seem like a big deal but big ships are slow and so these incidents often take place over many minutes or hours, thus there is plenty of time for somebody to wake up, realise there's a problem, wake someone else up to help fix it, and fix the problem before you smash the ship into an island clearly marked on your chart.
If Cloudflare has any sense they'll change the bogus route up every week or two so that outfits like A&A keep getting these "annoying" alarms and concerns from customers never cease.
> This is marginally better than routing to somewhere else (some attacker) but it still means a black hole in the Internet.
No. It's the difference between an inconvenience (routing problem which can be reported and then corrected by a human) and a grave security problem (data is being sent to completely the wrong place because you didn't bother with the mechanism to prevent that).
Also, this glosses over the _very common_ scenario where the bogus routes are more specific than legitimate ones. If AA doesn't have a route to one specific /24 containing secure.bankofamerica.com that has "somehow" been advertised by a Taiwanese ISP for the past hour, that data will be bundled along with the rest of the /16 or whatever else (BoA's assignment is old and a mess) to the Akamai boxes (I think) which actually serve it anyway. Rather than a "black hole" there was no problem except for the bogus advertisement, which RPKI would have helped block.
The choice to block one route to make isbgpsafeyet apparently pass is bogus, it's like silencing a fire alarm (which is also something I see far too many businesses choosing to do, oh Zone A keeps alarming, we'll just disable it even though now that means more danger, engineer calls are too expensive)
Those familiar with my work will know that I read a lot of accident investigation reports, and a recurring theme in maritime accidents involving large vessels is that the BNWAS (a technology which alerts sleeping crew elsewhere on a vessel that for some reason nobody on the bridge is paying attention, probably because they're asleep) is "somehow" muted, with its logging disabled or even entirely switched off during an incident. That might not seem like a big deal but big ships are slow and so these incidents often take place over many minutes or hours, thus there is plenty of time for somebody to wake up, realise there's a problem, wake someone else up to help fix it, and fix the problem before you smash the ship into an island clearly marked on your chart.
If Cloudflare has any sense they'll change the bogus route up every week or two so that outfits like A&A keep getting these "annoying" alarms and concerns from customers never cease.
> This is marginally better than routing to somewhere else (some attacker) but it still means a black hole in the Internet.
No. It's the difference between an inconvenience (routing problem which can be reported and then corrected by a human) and a grave security problem (data is being sent to completely the wrong place because you didn't bother with the mechanism to prevent that).
Also, this glosses over the _very common_ scenario where the bogus routes are more specific than legitimate ones. If AA doesn't have a route to one specific /24 containing secure.bankofamerica.com that has "somehow" been advertised by a Taiwanese ISP for the past hour, that data will be bundled along with the rest of the /16 or whatever else (BoA's assignment is old and a mess) to the Akamai boxes (I think) which actually serve it anyway. Rather than a "black hole" there was no problem except for the bogus advertisement, which RPKI would have helped block.