Re: [apops] Fwd: Korea Telecom leaking >1000 prefixes to Internet

  • To: Geoff Huston <gih at telstra dot net>
  • Subject: Re: [apops] Fwd: Korea Telecom leaking >1000 prefixes to Internet
  • From: Philip Smith <pfs at cisco dot com>
  • Date: Wed, 13 Dec 2000 05:36:34 +1000
  • Cc: Joe Abley <jabley at automagic dot org>, apops at lists dot apnic dot net
  • In-reply-to: <4.3.2.7.2.20001213052553.00c24100@localhost>
  • References: <5.0.2.1.2.20001213040422.07691610@localhost><20001212081205.B23618@goose.automagic.org><5.0.2.1.2.20001212141332.07842ec0@localhost><5.0.2.1.2.20001212042122.00ab3a40@localhost><5.0.2.1.2.20001212042122.00ab3a40@localhost><20001211135958.Q22842@goose.automagic.org><5.0.2.1.2.20001212141332.07842ec0@localhost>
  • Sender: owner-apops@lists.apnic.net
    • At 05:39 13/12/00 +1100, Geoff Huston wrote:
      > >Sean Doran made an interesting suggestion at the IEPG on Sunday. Charge 
      > providers for the prefixes they announce. So if anyone thinks that 99500 
      > is too much for their router to take, send a bill to the people who are 
      > announcing more than they ought to (i.e. subprefixes of their 
      > aggregates).  They may ignore it, or they may send a cheque... Who 
      > knows... Anyone want to test the theory?
      >
      >
      >The problem is identifying the difference between route push and route pull.
      >
      >You are implicitly assuming this unidirectional route advertisement heads 
      >'upstream' in making this statement.
      
      Upstream? I'm assuming that if someone makes a route announcement, it goes 
      pretty much everywhere in the Internet it is allowed to go in. So that is 
      upstream, downstream, between peers,... And if I understand the charging 
      point, you simply send a bill to the originator of the prefixes you'd 
      rather not carry in your network...? Or?
      
      >And who are the Routing Police and who gave them a charter and who gave 
      >them the authority to start intruding in the normally commercial in 
      >confidence relationships between a provider and a customer? The 
      >conversation strikes me as having identified a reasonable issue, but them 
      >wandered off into a vary unrealistic set of potential solutions. I _know_ 
      >that apnic would be __HORRIFIED__ to be portrayed as the Routing Police. I 
      >also know that the difference between collective action by a number of 
      >commercial entities and cartel-like behavior is often difficult to 
      >distinguish, and many providers would abhor the concept of behaving in a 
      >collusive way that discriminates or disadvantages a competitor or 
      >customer. So with all that in mind the solutions to issues such as these 
      >are not always easy to identify.
      
      Exactly, solutions are not easy to identify, and any kind of routing police 
      is not the right way to go. For starters, who would be the police?
      
      >The real issue here is far more complex and multi-dimensioned than is 
      >characterized in the messages I've seen to date, and before the routing 
      >jihad moves into the mode of eliminating the infidel /30 route prefix 
      >advertiser, we should all understand the dimension of the environment we 
      >live in, and its more than purely technology I would humbly suggest.
      
      Well, the particular issue I'm trying to "help" with is alerting providers 
      who are inadvertently announcing subprefixes of aggregates they are also 
      announcing. If KT, for example, have a commercial in-confidence reason for 
      announcing /30s etc to their peers, that is their private issue, but then 
      they should take reasonable steps to ensure that the rest of the world is 
      not drawn into their private affairs (the no-export community could help 
      with this, for example). But any provider who has a need to announce a /nn 
      to the Internet should be free to do so, as you say. And it should not 
      matter what it is - that's the idea of CIDR... :-)
      
      Some of the many examples I see, and know of, are ISPs who announce the /19 
      they received from APNIC, and then also announce the 32 "Class Cs" because 
      that they have been brought up on the historical class A/B/C system we used 
      to have in the Internet many years ago. And these providers in question 
      tend to have one or two links to other service providers. For one link, I 
      query why this leakage is necessary (the KT example falls into this 
      category, going by the AS PATHs). For two links, well, this is "multihoming 
      by the scatter gun approach". Nothing wrong with it, but it could be done a 
      little more optimally, and with a little more finesse?
      
      cheers,
      
      philip
      --
      
      *             APOPS: Asia Pacific Operations Forum              *
      * To unsubscribe: send "unsubscribe" to apops-request at apnic dot net *