- Test - lists.nanog.org

Re: The Attitude (was: the Internet Backbone)
by William Allen Simpson 16 Apr '96

16 Apr '96

> From: "Erik E. Fair" (Internet Architect) <fair(a)apple.com> > Perhaps the easy way out is to suggest that educating the ISPs as to > what constitutes good behavior at an exchange (routing system stability > and reliable packet delivery) is the responsibility of the exchange > operators, This is not just the easy way out, I believe that it is the only way out. Then, we don't have the problem of all those multiple peering agreements. Instead, just one with the exchange. Also, this eliminates the problem of each ISP trying to "filter" bogus routes. Instead, the exchange operators will handle that problem. Indeed, I think that this is/was a major impetus for "policy-based" routing in the first place (reading the old RFCs), and a clear reason we need the Routing Arbiter! > and it might even be possible to enforce some interesting > policies in that regard in the route servers (e.g. if you have more > than N routing or BGP peer transitions per time period, the route > server will refuse to peer with you for 48 hours - think of it as the > hold-down or damper from Hell). > Interesting concept! > I certainly think that to the extent that the exchange operators can > measure such things as routing and peer stability, it is in everyone's > interest to see the numbers (except those ISPs who are unstable). Who > knows? A series good reports from exchange operators about an ISP might > lead to offers of private peering arrangements outside of the exchange, > to the benefit of the ISP. Similar to the way that having a good credit > record seems to lead to endless offers of more credit. > I understood that the RA contract provided for these measurements, and agree that they would be a good thing! WSimpson(a)UMich.edu Key fingerprint = 17 40 5E 67 15 6F 31 26 DD 0D B9 9B 6A 15 2C 32 BSimpson(a)MorningStar.com Key fingerprint = 2E 07 23 03 C5 62 70 D3 59 B1 4F 5E 1D C2 C1 A2

3 2

Re: Some corrections
by Scott Huddle 16 Apr '96

16 Apr '96

At 9:59 AM 4/16/96, William Allen Simpson wrote: >Both Merit and CICnet are connected to MCI. Correct? My new >understanding is that the problem seems to be that MCI is stuck in the >middle there, and for some reason shunts some (but not all) OSU and >OARnet traffic off through MAE-East, rather than passing it more >directly to OARnet (through CICnet or otherwise). And this is because >of some _source_ based algorithm, since this only happens to Merit >sourced traffic, but not other CICnet traffic, as you are using MCI to >hold together pieces of CICnet? Sounds bizarre to me.... Bill, Oarnet is an MCI customer as well. they have a T3 connection to our network at Washington and another to the Chicago area. I think you might be confusing traffic to them over their dedicated line at Washington as going over MAE-East. Hope this clears things up, -scott (huddle(a)mci.net) MCI Internet Engineering

1 0

Re: Some corrections
by rrv＠uiuc.edu 16 Apr '96

16 Apr '96

At 9:59 AM 4/16/96, William Allen Simpson wrote: >Both Merit and CICnet are connected to MCI. Correct? My new >understanding is that the problem seems to be that MCI is stuck in the >middle there, and for some reason shunts some (but not all) OSU and >OARnet traffic off through MAE-East, rather than passing it more >directly to OARnet (through CICnet or otherwise). And this is because >of some _source_ based algorithm, since this only happens to Merit >sourced traffic, but not other CICnet traffic, as you are using MCI to >hold together pieces of CICnet? Sounds bizarre to me.... > >In any case, it still proves my point -- Merit and CICnet (and MCI) >somehow view each other as competitors and don't cooperate to carry >traffic by the "best" route, thereby clogging other areas of the >Internet (MAE-East) at others' unknowing expense. Bill, you are way off base.

1 0

Re: The Attitude (was: the Internet Backbone)
by William Allen Simpson 16 Apr '96

16 Apr '96

> From: Jeremy Porter <jerry(a)fc.net> > Bill, here is why you will never be taken seriously be > any ISP/NSP. If networks can't find ways to cooperaate without > prodecting there own interests, we all might as well give up > now. > I'll try to remember not to take myself seriously.... This is a most unusual argument: that we should cooperate without protecting our own interests; and if we cannot, we should give up. Are you sure that you know what you are writing? > THE ONLY WAY TO BUILD A SCALABLE INTERNET IS FOR PROVIDERS > TO BE MOTIVATED BY SELF INTEREST AND ECONOMIC FACTORS. > This is contrary to your above assertion. However, this does not match the historic record. A scalable Internet doesn't even require the concept of "providers". Nor does it require self interest. There are certainly other models that could be (and have been) followed. > This is exaclty an issue of customer education, If customers > wanted good connectity they CAN find it. > Really? Where? And who does the education? Using what data (as the providers themselves have refused to release it)? May I suggest that this form of argument was advanced for rather a lot of other fields, such as food, drugs, comparing doctors and lawyers and their fees, and purchasing of automobiles and refrigerators. As I remember, Consumers Union (I'm a lifetime member) was investigated at one time as a "communist conspiracy". This does not follow the subject of this thread, which was in regard to (lack of) cooperation with competitors. I don't see how customer education will affect the physical interconnections of carriers, except as perhaps encouraging consumers to sue their providers for fraud.... > Hm.. maybe Apple should have thought a little more > about their little T-1 link to a single provider when they > had about 3 T-1s worth of data to send. I offer to sell > more connectivity to Apple, but they WEREN'T interested. I question your objectivity. Sounds like you have a personal self-interest issue here. By my measurements, there was no problem with the Apple T1(s). The problem I documented was that Sprint didn't have enough bandwidth out of Texas (specifically to Chicago in my tests, but reportedly to other places as well) to ensure full T1 delivery for Apple while also supplying service to others in Texas. That is, Sprint was underprovisioned. Also, their links flapped a lot. > They didn't care that there application/distribution model was > broken, and breaking the net, they didn't want to fork > out the extra bucks, or deal with the internal politics > to put the release out on the West coast where they > had a much higher bandwidth connection. > I disagree, as Apple DID care very much! In fact, they eventually did put it out on other sites, including their west coast connection. How would it have helped if they had a T3 in Texas, when Sprint couldn't even handled another T1's worth of traffic to the same site? How would it have helped the net if they had been multihomed? You provide a separate link to Chicago, and load balancing between you and other providers? > I haven't seen many reports of Sprint problems since > Sprint put its policies in places and they had time work. Only recently has Sprint been cooperating with the RA enough to actually get some Sprint flapping statistics. I cannot help that you personally have not seen them.... Why not look at them yourself? > MCI on the other hand has been bleeding about 30% packets out of > San Francisco, on its OWN INTERAL network for several MONTHS. Too true. I did not mean to pick on Sprint exclusively, except where Sean seemed to make a claim that somehow the Sprint network was somehow operating better then others, and that it was due to his wonderful sagacity in refusing to carry others traffic. I have not (yet) heard the same non-cooperative attitude from MCI. WSimpson(a)UMich.edu Key fingerprint = 17 40 5E 67 15 6F 31 26 DD 0D B9 9B 6A 15 2C 32 BSimpson(a)MorningStar.com Key fingerprint = 2E 07 23 03 C5 62 70 D3 59 B1 4F 5E 1D C2 C1 A2

2 1

Re: Some corrections
by William Allen Simpson 16 Apr '96

16 Apr '96

> From: Dorian Kim <dorian(a)cic.net> > On Mon, 15 Apr 1996, William Allen Simpson wrote: > > I am admittedly confused here. CICnet provides connectivity from UMich > > to OSU, and the other Big Ten schools. CICnet is physically connected > > to UMich and MichNet, and also to OSU and OARnet. > > This used to be the case. This is right now not the case. This maybe the case > again in the future. Confused yet? Welcome to the world of academic > networking. :) >... > Depends on what you mean. AS266 traffic goes through MCI to get to AS1225. >... > I'm still curious where you got this false impression of the situation. > Apparently, I was missing some pieces (that had not been described in your earlier messages), and remembering "old" information (as above). Both Merit and CICnet are connected to MCI. Correct? My new understanding is that the problem seems to be that MCI is stuck in the middle there, and for some reason shunts some (but not all) OSU and OARnet traffic off through MAE-East, rather than passing it more directly to OARnet (through CICnet or otherwise). And this is because of some _source_ based algorithm, since this only happens to Merit sourced traffic, but not other CICnet traffic, as you are using MCI to hold together pieces of CICnet? Sounds bizarre to me.... In any case, it still proves my point -- Merit and CICnet (and MCI) somehow view each other as competitors and don't cooperate to carry traffic by the "best" route, thereby clogging other areas of the Internet (MAE-East) at others' unknowing expense. Indeed, this sounds even closer to the UK situation, where they are too cheap to install a local European link, and instead steal from the other vendors' links to Europe through the MAE-East commons. WSimpson(a)UMich.edu Key fingerprint = 17 40 5E 67 15 6F 31 26 DD 0D B9 9B 6A 15 2C 32 BSimpson(a)MorningStar.com Key fingerprint = 2E 07 23 03 C5 62 70 D3 59 B1 4F 5E 1D C2 C1 A2

1 0

35th IETF Report Available
by lscanlan＠cix.org 16 Apr '96

16 Apr '96

An extensive summary report of the 35th IETF, held in March in Los Angeles, California, is now available at the CIX web site, http://www.cix.org/Meetings/ietf-35.html Highlights of the meeting topics covered by Mark McFadden, CIXtra's technical reporter, include: Explorations of Alternative Routing and Topology IRAD Portable Site Prefixes Exchanges Stratum-based Aggregation Globally Redundant Routing Information DNS-based Logical Routing Conclusion Internet Architecture Board Character Sets in the Internet IAB, IRTF and IETF Issues Hearing on Returning a Standard to Draft Standard Status Applications and Transport Joint Session "Netscape Hack" HTTPng Caching T/TCP Transaction TCP HTTP 1.0 Short-term Solutions HTTP Version 1.1 HTTP and Multicasting Other Work Ipng Working Group IP Security IPv6 Mobility Routing Aspects of the IPv4 to IPv6 Transition Agents, Mobile Code and Intelligent Agents CIDRD Lou Scanlan Member Information and Services Email: helpdesk(a)cix.org Voice: 703 8 CIX CIX

1 0

Re: Some corrections
by William Allen Simpson 15 Apr '96

15 Apr '96

> From: Dorian Kim <dorian(a)cic.net> > > Likewise, a lot of traffic from Ann Arbor Michigan to Columbus Ohio > > travels via MAE-East, despite the fact that Merit is already connected > > to CICnet, which is in turn connected to Columbus (both OSU and OARnet). > > The problem is, Merit has no "bi-lateral peering" with CICnet. > > This is not exactly correct. CICNet Michigan(AS266) peers with Merit over a > FDDI ring in Ann Arbor. > > I suspect that the problem you are running into is that CICNet Primary > region (AS1225) which does not have physical connectivity to AS266 or Merit > provides connectivity for OSU, and peers with OARnet. > I am admittedly confused here. CICnet provides connectivity from UMich to OSU, and the other Big Ten schools. CICnet is physically connected to UMich and MichNet, and also to OSU and OARnet. Are you saying that despite the peering with Merit/MichNet and the traffic actually flowing to OSU, some other pieces of CICnet at the same location (OSU and OARnet) are physically and topologically discontiguous? That is, you don't carry your own internal traffic? > This problems has nothing to do with who has what bilaterals with whom, or who > is willing to peer with whomever else, but rather a simple constraint based > on topology. > > So please choose a more apropos example next time. Thanks. > The quoted "bi-lateral peering" was taken from your own private message on the subject. Perhaps I don't understand the use of the term "bi-lateral peering". I thought that it meant willing to forward all traffic from each party to the other party. Your private message indicated that you have such an agreement with OARnet, but not with Merit, and that it was Merit that is unwilling to make the agreement. I then checked with Merit (John Vollbrecht), who gave me a similar (though not identical) explanation. Topological contraints were not mentioned in your private message. Sorry if I misunderstood your message. WSimpson(a)UMich.edu Key fingerprint = 17 40 5E 67 15 6F 31 26 DD 0D B9 9B 6A 15 2C 32 BSimpson(a)MorningStar.com Key fingerprint = 2E 07 23 03 C5 62 70 D3 59 B1 4F 5E 1D C2 C1 A2

2 1

Re: The Attitude (was: the Internet Backbone)
by Justin W. Newton 15 Apr '96

15 Apr '96

At 10:50 AM 4/14/96 -0700, Internet Architect wrote: >Bill, > > A short answer to your refusal to peer objection in SPRINT's >policies in the form of a question. Which is better: > >1. peering with everyone at an exchange, including those ISPs who are >clearly clueless, and whose cluelessness leads to operational problems, >i.e. injection of bogus routes, black holes, routing loops, routing >flaps, and BGP peer transitions? Clue and size have very little to do with each other. I setup and ran a router at MAE east for a few months and am about to setup routers at MAE East and Pennsauken. To date I have /never/ blown someone else's routes into someone else unless I was intentionally readvertising those routes as per someone's request. My router was properly dampened, didn't cause routing loops etc etc. That is /not/ the basis of Sprint's policy. Sprint's policyu is based on the I'm big and you're not philosophy. If peering was based on properly making announcements and not flapping etc etc, sprint themselves wouldn't have a lot of peers right now. (Sprint has in the past advertised routes that were not theirs, and were in fact mine on numerous occasions). As for stability, well, things have gotten a lot better. > >2. qualifying each of your peers as being clueful, prior to peering >with them, such that you (and your customers!) don't suffer from the >cluelessness of others? Sounds like a great idea. Justin Newton * You have to change just to stay Internet Architect * caught up. Erol's Internet Services *

1 0

The Attitude (was: the Internet Backbone)
by William Allen Simpson 15 Apr '96

15 Apr '96

Catching up on my email after tax week (did anyone else notice that there were a lot fewer postings on this and other lists this past week?): > From: Sean Doran <smd(a)icp.net> > It's because I'm an evil asshole determined to protect my > employer's interests and make our shareholders rich. > > This is somewhat incompatible with protecting the interests > of our competitors and enriching their shareholders. > Gentlefolk, while I agree that Sean has been "mostly" right on many technical issues, I am seriously unhappy with this attitude! Nor is this attitude unique to Sean. Since my local ISP and NSP is Merit/MichNet, and I am an advocate of more regional exchanges, I have been trying to get them to interconnect with other local providers, so that local traffic wouldn't have to go through MAE-East (and suffer 30%-50% losses) to travel 3 blocks across town. They never managed to do anything in 3 years, despite the willingness of others (specifically MSEN) to interconnect. Recently, I asked why they aren't connected to the Detroit NAP. The response was "everyone else should connect to MichNet, and pay our affiliate fee". I noted that the others consider themselves competitors, and taking the same attitude would expect Merit to instead pay THEM, since MichNet generates the most traffic. Likewise, a lot of traffic from Ann Arbor Michigan to Columbus Ohio travels via MAE-East, despite the fact that Merit is already connected to CICnet, which is in turn connected to Columbus (both OSU and OARnet). The problem is, Merit has no "bi-lateral peering" with CICnet. Merit doesn't think there is a "cost benefit" to have regional interconnection and peering relationships. In some respects, they are right. The benefit is not to Merit itself, but rather to its customers (lower delays), and the rest of the Internet (less congestion at other exchanges). The problem is that ISPs are allowed to shove their regional connectivity out to others on the Internet. In effect, the _rest_ of the Internet is _paying_ for the regional underprovisioning. > Personally, I have little patience for the small and > not-very-clueful who want to be direct competitors with a > multibillion dollar company with lots of talent and who are > taken to whining about my policies and those of my > colleagues and associates, and even those of our > competitors. This uncharitable attitude obviously does not > endear me to them. > But your attitude that the whiners are "small and not-very-clueful" is less than useless. There are some quite clueful folks that don't agree with your policies, particularly with the failure to peer (and exchange traffic) with everyone else (even small folks) at an exchange. (Sprint is not the only perpetrator of this poor policy.) The fact is, whether you like it or not, they _ARE_ your competitors in their specific regions. But, to thrive, the Internet has a long tradition of _cooperation_ among competitors. Kinda misses the meaning of "exchange". That hurts everyone else on the net, by increased delay and more congestion elsewhere. In short, you are asking _others_ to bear the costs of _your_ making money. We've seen this time and again, such as the UK provider who sends all their traffic to the US, which then uses the congested US to Europe links. It only saves them money because others were unknowingly bearing the cost. Sounds like a form of fraud to me. > I would hope, though, that the bulk of our customers > would be much happier with us driving towards a network > reliable enough that they don't have to worry about their > customers screaming (not to mention not having to worry > about facing some very difficult scaling problems we are > already staring at), than with us being the Department of > Warm and Fuzzy Feelings. > > All of them know full well that the drive ain't easy. > True. But there are some particular bones to pick with Sprint, like the underprovisioned Texas links that kept dropping out, just when Apple released its 7.5.3 MacOS Update to developers from Texas.... So, let's see some of that vaunted reliability first, please. WSimpson(a)UMich.edu Key fingerprint = 17 40 5E 67 15 6F 31 26 DD 0D B9 9B 6A 15 2C 32 BSimpson(a)MorningStar.com Key fingerprint = 2E 07 23 03 C5 62 70 D3 59 B1 4F 5E 1D C2 C1 A2

5 4

Re: More on 'Number of Tiers' Discussion
by Michael Dillon 14 Apr '96

14 Apr '96

On Sun, 14 Apr 1996, Stan Barber wrote: > I didn't type it in. I just cut and pasted it. Here it the text of the original > note you sent: > Have a look at http://www.sidhe.memra.com/rough.txt if you want to see Mea culpa. This should read http://sidhe.memra.com/rough.txt Michael Dillon Voice: +1-604-546-8022 Memra Software Inc. Fax: +1-604-546-3049 http://www.memra.com E-mail: michael(a)memra.com

1 0