I think it's a matter of Cogent trying to strongarm its position. It wouldn't be the first time Cogent has done this and it certainly won't be the last. Doing a Google search for "peering dispute", and not including Comcast (to exclude the Comcast vs. Level3 dispute since it's newer and ongoing), almost every old entry involves Cogent duking it out with someone. They win customers on price, but things seem to be lopsided enough that they get into a scuffle with a number of the other Tier-1 providers.

Mike from HE spells it out pretty clearly from almost 2 years ago on the NANOG list:

http://www.merit.edu/mail.archives/nanog/msg01006.html

I have no reason to think that their stance has changed any.

Are you kidding? This isn't about merit. It's a game of odds.

This big law firm smells dollars, and lots of them. If they can squeeze any kind of settlement out of Zuck, it might be worth it just for their cut of the cash. They're in it for a big win, and for no other reason. This supposed email is what they'll hinge the whole case on.

There are actually reasons behind this. I've got a /29 from Charter Business myself, but this is why it is the way it is, based on my experience as a former Charter engineer.

In the days of old, customers were assigned their statics in WAN-side way as you describe. My parents used to have a static assigned to them from a WAN block on their CMTS. This was great because whatever allocation assigned was very efficiently used. Granted, this was back when nodes were combined 4:1 or greater on the small CMTS that was being used. A uBR7246 with 1x6 cards in the day could easily route traffic for over 48 cable nodes, at 2:1 combining on the upstreams, and 12:1 on the downstreams. (A whopping 150mbps for 48 nodes ... laugable today).

It wasn't all that long ago I remember some towns sharing a single downstream port. Now, enter node splits, and combining gets down to 1:1 in many cases. Even with a much larger CMTS (uBR10012 vs. uBR7246), it can't handle the same number of nodes. With redundancy failover switchboxes, there are only 35 downstreams per box (assuming 5x20 cards).

Now a problem exists as soon as the box's capacity is reached. If I need to split your node and move it to another CMTS to increase your available bandwidth, I need to coordinate with everyone who is moving who has a WAN side IP and tell them that their IP address is going to change on whatever date. This turns into an incredible shitstorm when one person stammers their feet and cries up the escalation chain and then delays necessary work because they bitch. Then capacity continues to be in hell until the move is finally approved. Then, there are the customers who ignore your voicemail and phone calls and then cry for a credit because they didn't pay attention until the move date.

So now what everyone is doing in order to make this easier is to assign you a /30 or /29 or whatever which you get from your modem. The modem sends that assignemnt up via RIP and it gets redistributed into the network. Now, it doesn't matter what town you're in or what CMTS you're on. Note splits and changes can essentially happen without you ever having to renumber your side. With the growing demands on bandwidth, it's not unheard of that you could move a couple of times per year, depending on the scope of the engineering changes.

Seems wasteful, but that's the sense behind it.

They cradle my balls, stroke the shaft, work the pipe, and drink the gravy.

It probably fell more into obscurity due to its horrific 2600 VCS port.

The port to the Atari 5200 was sweet and, at least by the screenshots, was faithful to the original. It was one of my favorite games growing up, in part because it wasn't some kind of terrible arcade knock-off.

I've seen what the RIAA and its ilk have been releasing lately ... and I'd argue that most of our records are worthless upon release.