Cuban’s Pipe Dreams?


In certain circles it is quite fashionable to mock the incumbents, the bells and the cable companies, and portray them as knuckle heads. I rarely make that mistake, for underestimating them is a big folly. Like overlords they protect the caliphate with ferocity befitting a sultan, and have survival skills that are unmatched.

Today, I read Mark Cuban’s latest missive, in which he proposes that incumbents should distinguish between local and Internet bandwidth. Why haven’t the done this, he asks, after all, he theorizes that bandwidth is far cheaper on a private network than on a public network. He proposes a cheap 200 megabits up and down local bandwidth connection, and premium prices if connecting to the Internet.

Mind you, at present you can get those kind of speeds via OC3 connections that run at about 155 megabits per second, but that a pretty pricey solution, and can go for $2000 a month. Clearly, not what the Dallas Mavericks owner had in mind.

Since his proposal sounded horribly like what used to be a circuit switched network, I wondered why did we spend all that money on packet switching. Still it was quite intriguing! I ran this by a gentleman who frankly knows more about networks than either me, and most certainly Mr. Cuban. “This is the same N squared problem; need n squared connections if you want to set up a network with N people,” he replied.

I begged him to spare us the technical mumbo-jumbo and break down the possibility, not reality of what Mr. Cuban was proposing. Was this even possible? Now this is a man who is consulted by some of the top players in the broadband business. His response was that at present this is not possible. “Not even with fiber to the home; the maximum there per household is around 50 mbps. Cablevision is working on offering 50 mbps using Narad product.”

A very senior executive at a mega-billion dollar company who has experience in building these type of networks told me that DOCSIS 3.0 would make 200 megabits quite possible, but its not anytime soon, not as point-to-point connections as advocated by Mr. Cuban and will cost billions. That $5 price point simply won’t work.

For now, Mr. Cuban’s idea is just a dream …. pipe dream.



In a unrelated note….$2000 FOR AN OC3! I can’t buy two bonded Ts for that here! Some one tell me that’s a typo, or just reinforce the logic that we should continue to throw everything we can at the last mile solutions, instead of actually lowering the cost of communications, or giving the telcos any real competition.

Dean Johnson

There wouldn’t have to be billing records on a per packet basis. Just a throttle between local and non-local. I have a Charter cable modem and it would be nice to do uncapped transfers to/from other local Charter cable modem customers, for whatever value of “local” deems appropriate. It would be extra fast while it was on Charter exclusive wires and slow down once it leaves them. It’s like not paying for calls that remain on your cellphone carriers own equipment, which is what Cuban was talking about.

Jesse Kopelman

In a way, regardless of pricing there is already a distinction between local and non-local in the case of people living in major cities. Access to multi-MB speeds, especially symetrical, is almost exclusively the province of city/university dwellers. If you want to engage in high bandwidth activies, like P2P and sending giant e-mails, you would prefer doing so with others who have high bandwidth connections. In many cases, those people are your urban/university neighbors.

Aswath Rao

If I recall correctly, in Honk Kong 100Mb service has a tiered pricing – unlimited transfer within the country and a cap on the international traffic.

I personally think the problem with Cuban proposal is more logistics and less technical. This essentially means that the access router has to vreate a billing record for each packet. We know that billing and collection are two expensive items in the telephony world.

Finally I am not sure whether the fundamental premise is correct: currently the access bandwidth is limited whereas the core bandwidth is almost unlimited.

Wes Felter

It is obvious that this idea presumes FTTH. According to Light Reading, in November the ILECs are expected to put out a RFP for GPON equipment running at 2.5Gbps with a 32:1 split (i.e. 32 customers sharing 2.5Gbps downstream). That’s at least 78Mbps per customer of raw capacity; the effective capacity is much higher due to statistical multiplexing.

I think Dean Johnson got to the essence of Cuban’s suggestion: “simply throttle connections … on the gateway between the local network and the internet” instead of on the last mile. IIRC this is conceptually similar to the way billing is done in Australia where local traffic is free and international traffic costs. What both of these have in common is that there is bandwidth monitoring/enforcement equipment needed at the border between the ISP network and the Internet. If such equipment is not in place today, it would have to be added and paid for.

I think some muni fiber networks already operate this way; it would be instructive to look at them and see if any of Cuban’s scenarios have emerged there. P2P file-sharing apps may benefit automatically from Cuban’s proposal, but what are they good for? :-) One of the great benefits of the Internet is the “death of distance”; will people be willing to go back to a two-tier local/long distance model?

Dean Johnson

First of all, I think you are too hooked on the particular bandwidth numbers. I think Mark is simply suggesting just much more bandwidth than is currently available. Rather than 200Mbit, gee wouldn’t a mere 60Mbit local connection be a delightful addition to your day compared to 3Mbit.

While it is undoubtedly more complex than I suspect, couldn’t you simply throttle connections, to the current speeds (ie. 3Mbit), on the gateway between the local network and the internet? The problem is insuring an aggregated total for the user, on external connections, matches the lower number.

This response to Cuban’s post exemplifies the typical “engineer missing the point” scenario where user requests shot down by superficial analysis and getting prematurely wound up in technical details.


Actually, Om, I think you misunderstood what Cuban is suggesting (or at least your consultant experts did). Cuban is merely suggesting that Telcos distinguish local access traffic from internet access traffic (i.e. on a logical level, not physically wire up point to point connections, as your consultant expert thought) and price them differently. I suspect he is suggesting something like Ethernet or FTTH. Sure, this kind of pricing model is something that the Telcos can definitely introduce once they have wired up the town; however the problem always has been the last mile and the huge capital costs involved in wiring up every home.

Comments are closed.