Hacker News new | past | comments | ask | show | jobs | submit login

This mangles two issues - discriminating based on source versus path.

Source discrimination is bad. The Internet allows applications and services to run “at the edge” of the network and not centrally; this encourages innovation [1].

Path discrimination is more complicated. Many content providers already pay private networks to transport their traffic on faster than the public internet [2]. There is even a market for traders paying tens to hundreds of thousands of dollars for low latency connections [3].

Given that ISPs charge each other for transporting traffic [4] it costs more to broker traffic across others' networks versus its own. It doesn't seem unfair for the ISP to charge less for the latter. This would allow Comcast et al to compete with the Akamais and Level3s that irk them today [5]. Sooner or later they will find it makes more sense to offer the discount for same-network traffic to everyone.

Comcast has made shitty statements about net neutrality before [1]. But it's not okay to vilify anything Comcast says by virtue of it being said by Comcast - that's straight up ad hominem.

[1] http://www.stern.nyu.edu/networks/Economides_Net_Neutrality....

[2] http://techcrunch.com/2010/11/11/level-3-lands-netflix-strea...

[3] http://www.highfrequencytraders.com/article/682/options-it-o...

[4] http://blog.teracomtraining.com/how-isps-connect-to-the-inte...

[5] http://blog.comcast.com/2010/11/comcast-comments-on-level-3....




Local Monopolies have little incentive to improve their infrastructure. When you enable them to charge premiums for degrading existing service to carve out 'bandwidth' for premium service everyone looses so they can make a higher profit from the same crappy service.

Bandwidth is cheap. The bandwidth to give everyone in a mid sized city 100mb internet access costs around 20$/person a month + the cost of the wires. So, if Comcast provides 5mb internet for X$ they can make money providing 100mb for X + 20$. Yet, they want to both charge 3+ times that AND make money on side deals AND keep their local monopoly.

Now, latency is a slightly different issue. I would suggest that it's reasonable to either run an extra line OR host some servers locally to deal with that. But, assuming you have a sane network topology there are vary few cases where latency is actually important. Yet, if Comcast sees an opportunity to profit from the latency game they will do so. They already provide crappy DNS service to slow people down just think what they will do if they think latency is the path to extra profit.

PS: You could replace Comcast with Cox and probably just about any other local cable company and say the same things. But, I just happen to know more about Comcast and they are in the article so I stick with them.


Where are you getting the $20 number from?

I do not want to be in the position of defending Comcast (who I hate with fiery passion), but that sounds a bit like arguing that the cost of manufacturing an Adobe Photoshop disc is relevant to how much it can or should cost.


Talking with friends in the that work at backbone companies. One of the funny stories I remember is the sales people having issues making sure that they actually charged more money for higher bandwidth connections due to some internal issues they almost set the prices for OC-12 lines (622 megabits) below that of OC-3 lines (155 megabits). As I understand it internally OC-3 lines where costing them slightly them more to deploy at the time.


Now, latency is a slightly different issue. I would suggest that it's reasonable to either run an extra line OR host some servers locally to deal with that.

How does running an extra line lower your latency? I think you're confusing bandwidth and latency here. Latency would involve replacing your line with better infrastructure (small impact) or changing the route your packets would tend to take to one with fewer hops and shorter distances (big impact).


If your latency is generally a combination of # of routers, physical path, and network congestion between you and the destination. Locally number of hops dominates the latency. So if the old path is comcast > Level 3 > google then running a comcast > google line will lower your latency by removing Level 3's routers and probably proving a shorter path.


Bandwidth isn't literally width like a pipe of water. It is linespeed. That is why higher quality cables have higher bandwidth, even though packet transmission is completely serial. "wdth" is an illusion created by time division multiplexing, like multitasking/multithreading on a PC.


This isn't always true because time isn't the only multiplexing strategy used. For instance, modern cable modems use code-division multiplexing, and fiber optic communications typically uses a variant of wavelength-division multiplexing. In both of the later cases, there is actually simultaneous transmission.


Level 3 was offering to transport the data all the way to Comcast's "internal" network and willing to pay for the equipment required for the high-speed interconnect. How can Comcast complain about not having enough bandwidth for delivery in their last mile (or least need compensation to build out more bandwidth), but then turn around and offer a very similar service and not even have it count against their bandwidth caps? Is there a bandwidth issue in their internal network or not?

I doubt Comcast put in data caps because of interconnect costs. I would love to see what a gig of interconnect data costs them.


A September 2008 price sheet for Level 3 priced 10 Gbps at $6/Mbps [1]. The same group projects 2012 transit prices at the minimum commitment around $2.34/Mbps and decreasing at about 60% a year [2]. Another source pegs this at 30% a year for the single price blended rate [3].

In the US we pay about $3/Mbps [4], suggesting a healthy minimum 1/5 profit margin. This happens to be around CMCSA's 1/5 operating margin (suggesting the transit price is lower) [5].

I couldn't find data for internal transit prices, but assuming everyone runs with about a 1/5 margin on transit costs as well (likely too dear) we can take $1.9/Mbps as being a decent relative estimate of internal costs. This is a $0.45/Mbps or 15 percentage point to revenue difference.

[1] http://drpeering.net/AskDrPeering/blog/articles/Peering_vs_T...

[2] http://drpeering.net/white-papers/Internet-Transit-Pricing-H...

[3] http://conferences.sigcomm.org/sigcomm/2011/papers/sigcomm/p...

[4] http://dailyinfographic.com/internet-speeds-around-the-world...

[5] http://www.google.com/finance?q=NASDAQ%3ACMCSA


Thanks for the links. The problem with your profit margin calculation is it assumes 100% of consumer traffic is going out to a different network and that Tier 1 ISPs have arrangements where incoming and outgoing traffic on their networks cancel out so they usually end up paying each other very little money.


Yup, that's why I labelled it as a minimum profit margin and parenthetically noted that the cost per Mbps used is probably higher than reality.


What's the duty cycle on those cost numbers? A 250 GB monthly data cap would be 0.8 mbps at 100% duty cycle, which suggests that the interconnect bandwidth costs are a very small fraction of the price of a 20 mbps plan.


it's total bullshit on Comcast' part. If you remember, the argument was that the top 5% were causing issues for the remaining 95%, and the issues were stated, at the time, to be with the 'last mile' pipe. Just another example of a monopoly being a monopoly.


>There is even a market for traders paying tens to hundreds of thousands of dollars for l

That's not for consumer broadband access. It's not even for commercial broadband access. It's a specific use, point-to-point network connection. It's important to separate general broadband access from special-purpose network connections for net neutrality purposes.

Paying a company to bring fiber into your area and then leveraging their service over that fiber is very different from consumers dealing with (typically) municipal contractually-enforced monopolies that restrict their access to the internet. I'm a capitalist kind of a guy, but monopolies destroy the dynamics of the free market and have to be kept firmly in check.

>Given that ISPs charge each other for transporting traffic [4] it costs more to broker traffi[...]

I don't think anyone is arguing that there isn't a cost motive for providing tiered pricing. Of course there is.

The issue is that monopolies must either be broken up or they must provide special protections for both consumers and the health of the free market in general. Net neutrality is that protection and it trumps Comcast's cost-based-pricing motives.


So if Netflix can already buy bandwidth on Comcast's public network, why can't I stream Netflix without it counting towards my bandwidth cap? Same with hulu or any other video service. If Comcast were offering this service to anyone, I could see your argument being valid. But until that point in time, I'm going to say that this is anti-competitive.


So, what you're saying is that any ISP could cache whatever they wanted from the web (major sites), offer it unlimited on their own network and then cap/limit/charge more for the rest of the web. What could go wrong with that?


I agree. The article lost me at this bit: 'The company claims that “content is being delivered over our private IP network and not the public Internet,” but that’s a meaningless distinction given that Comcast owns the entire pipe and can arbitrarily create private networks any time it’s convenient to do so.'

What does the author mean? Comcast certainly doesn't own the entire Internet, so why is it a meaningless distinction?


The public Internet is a bunch of connected private networks with peering agreements. So comcast can flip the public bit by turning off peering whenever they want.


Making that distinction would mean Comcast is justified in pricing it differently, but since we already know Comcast is not justified, the distinction must be meaningless.


I agree, but what happens when Comcast then buys out all the content providers?




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: