Sprint re-enabled the connection between Sprint and Cogent at 21:00 UTC (16:00 EST) on Sunday, 2 Nov, 2008. Sprint issued a hastily prepared statement about the reconnection (the HTML is a cut-and-paste job from "IP/MPLS Products from Sprint"), explaining their position. Cogent hasn't commented yet.
The connection appears to be routed much as it was before Oct 30. Previously, we saw Sprint selecting 2700-2900 prefixes from Cogent (that is, picking Cogent as the best path for that many network prefixes). We saw Cogent selecting about 7500-8000 prefixes from Sprint. Now that they have reconnected, Sprint is selecting 2538 prefixes from Cogent and Cogent is selecting 7016 from Sprint. So down slightly, but not appreciably. The link is up.
The fact that Sprint has reconnected this indicates clearly that they intend to fight this battle in court rather than in the routing tables or in the court of public opinion. This fact alone makes this likely to be one of the more interesting peering disputes of the last few years. But the resolution may take months or years, given the speed with which the courts move.
If they plan to resolve this in court, why disconnect in the first place? Do you think that Cogent's response (or lack thereof) forced them to change plans, or was the depeering just a warning shot?
It's hard to guess what will be the eventual solution. Is it Cogent or is it Sprint that stands to lose the most? An interesting third way may be that both networks will have to buy transit from a third party again at which point the costs of having active peering sessions may be significantly less than the costs of having the transit link in place.
I do get the feeling that being a Tier1 may be less preferable than some parties think. Tier2 (with plenty peerings) offers many of the benefits and a fall-back to transit if all else fails (either deliberately or by accident)
For some more explaing of peering and transit have a look here: http://arstechnica.com/guides/other/peering-and-transit.ars :-)
Sprint claims that they've been disconnecting ports two per month for the better part of a year, with no response from Cogent and that Thursday was simply the occasion of the last two ports. My guess (not being a lawyer, or working at Sprint or working at Cogent) is that Sprint wanted to show actual harm and that that harm would be caused by Cogent's unwillingness to pay NTT Communications (AS2914) for transit to Sprint. There is now not just a billing dispute (which can be solved by financial compensation) but also a matter of specific performance on the table (demanding that Cogent route to Sprint via other carriers, or not).
Cogent's version of the story is pretty simple: they paid port fees for 6 months during a 'peering trial'. They met the criteria. They stopped paying the fees. Then Sprint unilaterally changed the peering criteria 6 months later, and Cogent failed to meet them.
There's an interesting question about the metrics. Cogent says that the peering stat was supposed to be measured at 95th percentile (which they refer to as "industry standard", despite themselves offering billing at 90th percentile :-) ). Sprint then switched the metric to 'average' which Cogent says disadvantaged them and violated the agreement. Now, naively, I would have thought that average would benefit Cogent more than 95th percentile (unless they were maniupulating the routing on a monthly basis to try to push up the inbound Sprint->Cogent stats for >36 hours every month), so this may require more thought and more information.
@ ToddIt all depends on what the average refers to. If Sprint first wanted to see the 95% reach a certain threshold x and they have now changed that to the average needing to reach x, than that is a serious increase in threshold.
What I do wonder about is why Sprint is holding on to its position so strongly. It isn't like they will see an increase in income even if they win this battle. Cogent might be forced to buy transit somewhere, but it will not be from Sprint. Paid Peering would weaken its position. Sprint would lose quality in its network, but given its zombie status might not care to much. It doesn't sound like the upkeep of the peering was costing them too much money, so why do this? Where is the money? (Did NTT buy transit from Sprint? If that is the case, then at 5 dollars per mbit/s/month I would estimate the lost income at about $100.000 a month for both NTT and Sprint.)