[187212] in North American Network Operators' Group
Re: The IPv6 Travesty that is Cogent's refusal to peer Hurricane
daemon@ATHENA.MIT.EDU (Matthew D. Hardeman)
Fri Jan 22 20:28:05 2016
X-Original-To: nanog@nanog.org
From: "Matthew D. Hardeman" <mhardeman@ipifony.com>
In-Reply-To: <CAP-guGV2XdoxtCHfoKNXzjJaKnZ1Jw4W2zLmZyykfPfBOyufTw@mail.gmail.com>
Date: Fri, 22 Jan 2016 19:27:57 -0600
To: William Herrin <bill@herrin.us>
Cc: "nanog@nanog.org" <nanog@nanog.org>
Errors-To: nanog-bounces@nanog.org
Bill,
I find that I agree with much of what you=E2=80=99ve said.
If we further constrain the arguments that you set forth so as to cover =
only that traffic which the customers of the two networks would be able =
to exchange in any event, by way of transit services purchased by one or =
the other of the two networks, then I agree wholeheartedly, at least on =
a purely logical basis.
In that instance, the traffic is exchanged regardless (though often over =
links that saturate at peaks) and furthermore at additional expense to =
one or both of the networks involved.
=46rom a logical perspective, if two networks will permit their =
subscribers to exchange data, why would those two networks not elect the =
least cost, highest quality mechanism for exchanging that traffic?
I can only think of economic reasons, and specifically the hope for =
potential revenue from the other networks=E2=80=99 customer, because the =
parties have been unable to exchange data reliably over congested =
transit links.
Look, for example, to what was quite obviously the intentional =
peak-period congestion on various Comcast transit and peering links.
I=E2=80=99ve personally acted in a technical and administrative capacity =
in helping clients of mine (voice service providers) add private paid =
peering / paid customer links into Comcast just to overcome voice =
quality issues during peak periods resulting from clearly congested =
transit and peering links. It was obvious during those arrangements =
that Comcast had chosen to allow those links to congest as a policy =
matter in order to extract additional revenue by charging desperate =
=E2=80=9Cnew customers=E2=80=9D a premium toll for access to their =
subscribers behind the wall-of-congestion.
What=E2=80=99s fundamentally different in this IPv6 only Hurricane =
Electric <-> Cogent matter is that rather than have the traffic flow via =
transit (whether congested or not), there is quite simply no path =
between those two IPv6 networks. Hurricane Electric, clearly the IPv6 =
leader refuses to engage in the purchase of transit services for IPv6, =
and Cogent refuses to peer with HE on either protocol no matter what. =
Thus, no flow of traffic between the two networks on IPv6.
Presumably Cogent=E2=80=99s policy is mostly about denying Hurricane =
Electric to the =E2=80=9CTier 1=E2=80=9D club, on IPv6 that ship has =
sailed. Let=E2=80=99s face it: when the really tough Tier 1s are =
peering with you (like Sprint, Level 3, AT&T), you=E2=80=99re in. Even =
Sprint peers with HE on IPv6 (though they do not on IPv4). Honestly, I =
think Cogent is the only hold-out. At least the only one that matters.
In as far as HE maintains an open peering policy both for IPv4 and IPv6, =
it=E2=80=99s clear that Cogent is the bad actor, denying their customers =
a path to Hurricane Electric customers. I think the only reason this =
has been tolerated so far is that IPv6 has been a fringe matter until =
now. Even today it=E2=80=99s a minority of network traffic, but it=E2=80=99=
s gaining fast.
If I were Cogent, I=E2=80=99d be more worried about denying my customers =
access to HE=E2=80=99s IPv6 network than the other way around.
Matt Hardeman
> On Jan 22, 2016, at 7:03 PM, William Herrin <bill@herrin.us> wrote:
>=20
> On Thu, Jan 21, 2016 at 1:52 PM, Brandon Butterworth
> <brandon@rd.bbc.co.uk> wrote:
>> I'd like to peer with all tier 1's, they are thus all bad as
>> they won't.
>=20
> Correct.
>=20
> I've said it before and I'll say it again: an ISP's refusal to
> maintain a settlement-free open peering policy is directly linked with
> said company's fraudulent double-billing for services.
>=20
> In case you don't see it, I'll explain: whatever fictions you may tell
> yourselves, your customers pay you to connect them to the entire
> Internet. So do the other guy's customers. Settlement free peering
> means that at no _additional_ charge to anyone, you accept the packets
> your customers have paid you to accept from the other guy's customers.
> And vice versa. Peering does not trade packets you haven't been paid
> for. That's another fiction. Peering only trades packets one of your
> customers has paid you for.
>=20
> I get from there to double-billing because the alternative to
> settlement free peering is a paid relationship. The other guy has to
> buy from you directly (becoming the second payer for each packet) or
> he has to buy from one of the peers you've accepted But the peers
> you've accepted are constrained by ratios an related technical
> requirements which functionally prevent them from adding a sizable
> amount of traffic from that other guy, so unless he's doing a trifling
> business he pretty much has to buy service from you. Even though
> another customer has already paid you to perform that activity, you
> refuse to do the job unless the second party also becomes your
> customer and pays you. Fraud. Hidden behind a wall of technical
> minutiae but fraud all the same.
>=20
>=20
> Don't get me wrong. You can cure this fraud without going to extremes.
> An open peering policy doesn't require you to buy hardware for the
> other guy's convenience. Let him reimburse you or procure the hardware
> you spec out if he wants to peer. Nor do you have to extend your
> network to a location convenient for the other guy. Pick neutral
> locations where you're willing to peer and let the other guy build to
> them or pay you to build from there to him. Nor does an open peering
> policy require you to give the other guy a free ride on your
> international backbone: you can swap packets for just the regions of
> your network in which he's willing to establish a connection. But not
> ratios and traffic minimums -- those are not egalitarian, they're
> designed only to exclude the powerless.
>=20
> Taken in this context, the Cogent/HE IPv6 peering spat is very simple:
> Cogent is -the- bad actor. 100%.
>=20
> Regards,
> Bill Herrin
>=20
>=20
> --=20
> William Herrin ................ herrin@dirtside.com bill@herrin.us
> Owner, Dirtside Systems ......... Web: <http://www.dirtside.com/>
