"Finally, there's the question of metered billing for ISP customers." This has nothing to do with net neutrality.

Actually, it can. Cable companies offer both a video service and an internet service. They have a very strong financial incentive to limit the usage of Internet-based video services (e.g. Hulu), since they cut into the need to use the cable company's own video service. The solution is simple: impose bandwidth caps, regardless of the actual marginal cost of providing more bandwidth.

I'm not saying that they can't impose caps, but they need to be reasonable. Time Warner's recent attempt at caps was pretty bad -- it was clear to me that they were trying to protect their video business -- while Comcast's 250GB seems better.

I think this report is misleading. It should be titled "97 of top 100 pure-play classifieds sites are craigslist".

Hitwise assigns every web site to a single category, like "classifieds" or "news". If your site spans multiple categories, then you have to choose the one that defines you best. Since many traditional classifieds providers are also large news sites, you'll find that there are a lot of major sites that are missing from this list of "classifieds" sites.

I'm not saying that craigslist isn't the powerhouse of online classifieds, but to simply ignore a wide range of classifieds providers because they also provide news is kind of silly.

This isn't interstate commerce. Telling a company, "You can't create a website at www.gambling.state.ky" is no different than telling a company they can't build a gambling house in Louisville.

Bad example, since they are seizing .com domains, not .ky (though you probably meant .ky.us). Perhaps if the gambling sites had some physical presence in the state (e.g. servers or employees) then Kentucky could try to shut them down, but most of those gambling sites aren't physically anywhere close to Kentucky.

Go visit Utah sometime. If you want gambling or porn, you have to drive to Nevada.

You're only half right -- Utah doesn't allow gambling, but you can buy Playboy in a store. My wife used to work at a major bookstore chain which carried them.

That's what your cable bill is for -- to pay for the wiring and access to the programming.

Not true:

According to research firm SNL Kagan, Viacom's MTV receives about 32 cents per subscriber per month from cable operators, while Nickelodeon receives an average of 45 cents a month. Operators pay 86 cents a month per subscriber for the Disney Channel, which commands a higher fee because it runs commercial-free.

It's not marketing-speak.

You bet it's marketing speak -- you didn't quote any numbers, instead preferring to use the term "enterprise-level".

Hint: small frames/packets can often kill commodity PC routers.

Perhaps you should drop that hint to Cisco -- most of their small to mid-sized routers have the same exact issue. It isn't an issue that's limited to commodity PCs.

And to be clear, Vyatta might very well be able to do 2-3 Gbps with 64 byte packets.

Of course it's with 64-byte packets: that's the common lingo of network hardware manufacturers. You can find similar throughput measurements on every piece of Cisco or Juniper equipment. Anyone that quotes bandwidth throughput in passing will use the 64-byte figure, since it's always the highest one.

No, not possible. GigE is hardly relevant anymore.

Huh? Gigabit ethernet is "hardly relevant"? What world are you living on?

Even so, you can't get enterprise-level packets per second performance on commodity PC hardware.

Well, it's hard to refute a statement that uses marketing-speak like "enterprise-level pps performance". A commodity PC can achieve gigabit throughput, though: Vyatta claims their x86-based 2502 appliance can achieve 2-3 Gbps.

GigE kit is cheap enough to not do it yourself.

I guess it depends on how you define "cheap".

This you miss the part where I pointed out that they have half a million servers?

There is not a direct correlation between number of servers and bandwidth usage. In addition, Google wouldn't necessarily build a router that satisfies every routing need that they have -- they would build a router that would effectively and cheaply satisfy some of their needs. I have little doubt that, even with their own commodity routing platform, they would continue to buy high-end routing platforms for certain things.

Plus, you shouldn't discount the performance of a server-based router. The NANOG mailing list recently discussed the topic, and at least one person claimed to have sustained 970 Mbps throughput on a low-end Dell server. Given the deep technical expertise at Google, I imagine they could get it well above that, too.

The point I was rebutting was the assertion that Google might sell an external version of the router, because they were already selling an external version of the search engine.

The point is that Google already has experience selling a server-based appliance. That experience would probably be very applicable to selling a server-based routing platform, at least in terms of the inventory management, order fulfillment, shipping, etc...

I'm not an expert on this, but it's my understanding that a cost-efficient router is very different from a commodity system.

Not necessarily. A Linux computer with the correct commodity hardware and configuration (PCI Express interfaces, MSI interrupts, etc...) can easily route hundreds of megabits per second of IP packets. Of course, the PC only works well for certain applications, such as forwarding between a few low-cost interfaces with mainstream BGP/RIP/OSPF routing protocol support. This essentially challenges the low- to mid-end of Cisco (i.e. the 7200 series router and lower).

Once you begin to get into higher port densities, esoteric protocol support, very specialized configuration options, or very high throughput with consistently low latencies, you'll need to look at the specialized hardware from Cisco or Juniper.

Their search appliance bears no resemblance to the systems they use in-house.

Of course it doesn't, but that's not the point.

I think you're mostly right: Google will never sell specialized router hardware.

However, I doubt that Google uses special router hardware even for themselves. I'd bet that a Google router platform would be based on a commodity PC with a few PCI Express gigabit ethernet adapters, installed with open-source routing software. Google likely has no interest in supporting the billion weird or legacy options that are present in the Juniper and Cisco products, so it's able to make a commodity unit that is a tenth the cost.

They also may consider selling it, just like they sell their Google search boxes. If it's all based on commodity parts, they could just team up with a server manufacturer and provide an installation disk with their software. The manufacturer would already have everything in-place to manage inventory and shipping, and Google could just focus on the software end of things.

Personally, I would love to buy a PC-based Google router. I don't need to get a Cisco device that costs $10k-$20k just to do forwarding, BGP, and IPv6, and having the Google name behind a cheaper box would be easier to sell to management.

NAT is harmless to any application that is not broken in the first place. There is never justification for putting an IP address inside the application layer.

That's a rather silly thing to say. I would agree that applications should avoid handling IP addresses directly in their application-layer data, but to say that it's never justified is just stupid.

In fact, it was a mistake to have applications resolve DNS; that should be a function of TCP/IP itself.

You apparently don't understand the concept of layering.

Added to the fact that Office's volume licensing makes it much cheaper than what you'll see if you poke Amazon.com, and in time of recession, its the LAST suite of apps that will be switched over...

You're mostly right, but OpenOffice can be used in two different ways: as an office suite and as a weapon in negotiations. Most large companies will do exactly what you say they will -- they'll stick with Office no matter what -- but there will be some that tell Microsoft to lower their price or else, even if they have no real intention of switching.

It's that gradual erosion of Office prices that truly threatens Microsoft over the next 5-7 years, not sudden large defections to OpenOffice. The huge margin that Microsoft has on Office is one of the two financial pillars of the company, and anything that jeopardizes that is a major threat to the company.

The sad truth is that MS is doing better than ever - they've more or less successfully diversified and have multiple profit centres.

That's somewhat true. Here are some operating income figures from their 2008 annual report -- it's 6 months out of date, but represents the most recent full year of data:

• $13 billion from Client (Desktop Windows)
• $12 billion from Business (Office)
• $4.5 billion from Server
• $0.4 billion from Entertainment (Xbox/Zune/PC games)
• -$1.2 billion from Online Services

My take on it is that they aren't diversified enough -- everything hinges on desktop Windows and Office right now. There's some strength in the Server division, but that's also where they have some very powerful competitors. If I were an investor, I would pay close attention to corporate spending in 2009, since some companies may start exploring cheaper alternatives.

The Entertainment division is definitely a weak point, especially when you consider that the Xbox is approaching its peak profitability. Even if they were making ten times as much, I would still be wary of depending on the Xbox's revenues, since the market leader can flip-flop between console generations.

In the US companies that deliver goods or services for the public are generally perceived to be protected from litigation since they are considered to be a "common carrier," but this is only the case if they do not discriminate with what things they traffic.

Wrong answer. Facebook has every right in the world to delete content from their site. Research the CDA and Zeran v. AOL.

Would laws like this regarding breastfeeding translate into the online world?

I certainly hope not. The restaurant essentially deprived the child of food, while Facebook simply wouldn't let a woman exhibit her photos on their site.

Also keep in mind that Facebook does allow photos of breastfeeding, so long as the nipple and areola are not visible. It seems to me that Facebook is trying to appeal to the broadest set of users possible, and if they feel that female nipples hinder that, then more power to them. Those people that don't like it can set up their own site and stick the photos there.

Once, a long time ago, people wrote code in assembly... Of course cheap powerful hardware has made that all a thing of the past.

Actually, I would argue that advances in compilers and interpreters have been just as important to that trend as advances in hardware.