Myth 1: The Internet can be “neutral” towards all types of applications

Status: true! The Internet can be and usually is neutral towards applications. No matter which ISP you use, most of the Internet is someone else, and although your immediate ISP may have particular policies in relation to specific applications, you can be pretty sure that every set of intermediate routers between you and any given random end-point won’t share that policy. Neutrality is the rule, not the exception, simply because the network is so diverse. Rules are expressed as divergences from neutrality.

Some applications have different preferences as to the kind of service they would like, and IP was originally designed with a “Type of Service” field to address this. Note that placing any particular value in this field has never guaranteed that a router would handle it differently—it’s just another data point on which a router can base its handling, should it choose to do so. Since then, there have been standardisation efforts towards redefining this field for “Differentiated Services” [RFC 2474, RFC 3168]—a more complex embodiment of the same basic idea—and also “Explicit Congestion Notification” [RFC 3168].

Ultimately, however, applications should not rely on the network to provide any kind of preferential service. The Internet is a “best effort” delivery system, and applications must be designed on that basis.

Myth 2: Network management is unfair

Status: false! Network management is not unfair in and of itself. It is easy to devise extremely unfair policies and apply them to a network, but it’s also possible to facilitate fair allocation of resources this way, given that some participants might be greedy. “Fairness” is a pretty nebulous term, though, and there are a lot of differences of opinion as to what constitutes “fair”. In a commercial environment, where service is being bought and sold, “fair” has connotations of “fair trading”, meaning that “fair” can depend a great deal on what the customer reasonably expects to obtain in the way of service.

Myth 3: Network management violates privacy

Status: false! Like Myth #2 above, network management does not violate privacy in and of itself. That’s not to say that it can’t. If you are responsible for managing a network in a way that does not violate privacy, you must be at least a little careful in your choice of mechanisms. After all, there’s a lot of data flying about that could easily be used to violate privacy, particularly given the ISP’s knowledge of the subscriber’s actual identity.

Myth 4: DPI is just a P2P “Throttling” Technology

Status: false! It should come as no surprise that Deep Packet Inspection has more than one use. The notorious “Great Firewall of China” reputedly uses it, for example. Many network diagnostic tools also make use of it. There are also arguments against using it in particular contexts: for example, routers which base their behaviour on the specifics of higher protocol layers make for a brittle, unpredictable, and ultimately unmanageable network. If a router needs to know something about a packet, that data should be in the IP header.

Overall, DPI provides the tools necessary to manage the network to ensure fairness, reduce costs and optimize revenues.

Yes, true for the operator. But what is fair? The operator can reduce costs by limiting bandwith for ‘unwanted’ traffic such as bittorrent (because it use much capacity) or Skype (because it competes with the operators services), but is this ‘fair’?

Allowing operators shape traffic is decidedly good for the operator, as it is in control, and possibly a disaster for a free and open Internet. If DPI had been widely deployed and unregulated new applications such as bittorent, Skype etc would have been killed in its infancy.

What adult expects the world to be fair?  It’s not, never has been, never will be. The Internet is no exception.  The Internet is not (no longer) a Government subsidized service provided for free.  When it was run by the Gov’t, it didn’t benefit the masses like it has since it was commercialized.  Not even close.  The Internet is run by commercial companies who invest their own (or their shareholders) dollars in infrastructure and service offerings.  If it were your investment, what would you want them to do?  To a degree, they are limited by technology and the price points that are acceptable in the retail market.  They need a return on the investment, else why do it.  Services are derived around those restrictions, as well as the laws that play into it.  There has to be some rules, policies and limitations, and some are passed on to the subscriber base (if not, they get passed on as costs somewhere.)  Every technology that has ever been commercially deployed has it’s technical or business-related limitations.  Even if the service providers do a poor job of alerting customers to how their service works and what the restrictions are, no one necessarily has an absolute right to do whatever they want.  BitTorrent and other P2P apps wreak havoc with broadband service, and a few users ruin it for the majority who aren’t doing P2P.

Look at it this way (apologies in advance to those who recently said how much they hate simplifying analogies), if you rent a car and the model you get can theoretically go 150mph, does that imply that the rental car company is giving you express permission to (a.) break the law, (b.) endanger yourself or others around you and (c.) risk damage to their asset?  Whether they express it or not in the rental contract, there are always limitations whether stated (well or poorly), and common sense and compromise have to play a part.

Good job on Brett Watson for seeing it right!

Myth #1 is obviously TRUE and has been through most of the history of the Internet.  Only recently did hardware makers came up with Service-Provider scale DPI devices cheap enough for large ISPs to use them, and then they started “shaping” the network claiming how necessary it all is.  Well we’ve had great success in the Internet without DPI.

And shame on the article for making up Myth’s #2, #3, and #4 all by itself and then debunking them as if it just proved something.  Nobody thinks that network management is unnecessary, nobody thinks that network management itself violates privacy, or that DPI is just good for throttling.

And I do wish more of the “NN Squad” types actually knew what they were talking about, or knew packets from a hole in the ground for that matter…

1. Yes there are patently unfair or even wrong ways in which DPI, per application traffic shaping etc can be used.

2. It is perfectly possible to use them fairly, to provide an acceptable user experience to all your customers (and say, prevent 1% of your users from leeching 90% of your bandwidth).

3. All you can eat / unlimited broadband connections are dumb marketing, because you bank on pipes remaining underutilized.  Try getting 2 mbit guaranteed pipes (aka a full T1 instead of a cheap ADSL) and see how much that costs?

That’s well on the way to becoming one of the classic logical fallacies.

thanks
srs

The email spam analogy is interesting but ironically can work both ways.  I know what you are saying, but spam has become such a problem that ISPs and business are using invasive tactics to thwart it.  Spam filters often cause collateral damage to legitimate email, which at best might end up in your junk mail folder or worse just never arrive.  But your point is not lost.  I merely wanted to point out the root of the issue because I see a lot of disinformation spread about legitimate P2P, which right now is not the majority.  Just because the tactics used to thwart an issue are questionable, illegal or deserving of criticism, it doesn’t absolve the root cause of the problem.  If it were as easy as it is to duplicate and trade music, software and movies as it is to photocopy books and other copyright-protected material, maybe the courts would have taken a stronger stance 30 (??) years ago on Xerox and photocopiers.  But the difficulty in standing at a copier photocopying each page of 2-inch-thick 1000-page Steven King novel was its own discipline!  In the digital age, it is too easy.

When companies like Sandvine and P-Cube developed their products, it was in direct response to Napster, BitTorrent, KaZaa, etc., a clear market demand.  When Cisco, the biggest network vendor of all and one who often acquires companies, decided to buy their way into the market (which resulted in the P-Cube acquisition 3 or 4 years ago), they weren’t sitting around a boardroom speculating on some future need; they were dealing with immediate demand because of media file sharing.

If the whole Comcast/FCC thing leads to usage-based services – and some ISPs are already testing this model - there will be a lot of collateral damage to innocent bystanders in the form of bigger bills.  At a minimum it will detract from the user’s Internet experience when the amount of time they spend online and the size of data they download preys on their mind.  It would probably stifle innovation as well, for similar reasons.

Major ISPs like Comcast will probably tread lightly here, knowing the backlash that will occur given that this has already played out in other services.  The backlash would first be a significant increase of calls to their call center, with subscribers first angry only about their bills but then becoming increasingly agitated as they find themselves on hold for an hour, a la the AOL scandal in 1997.  Eventually it will cause churn, and the bigger ISPs know this.

What may also happen is that someone will create a freeware application you can load on your PC that measures your traffic and creates a simple end-of-month report that you can use to challenge your Internet bill, pointing out the unsolicited traffic your Internet link received (port scans, attacks, etc.) or viruses that generated outbound traffic and should have been taken care of by the ISPs anti-spam filter.  Your ISP may say you owe X but you can say no, I didn’t ask for that traffic, I should only have to pay Y.  The application could even be clever enough to compare the size of downloaded files with the actual amount of data transmitted and, if there were retransmissions because of network errors or congestion, suggest to the ISP that that is not the subscriber’s problem either and we aren’t paying for the extra traffic.  So the arguments would continue.  ISPs would have to find that area between the majority of their subscriber base and the download junkies and fix the usage-maximum there to penalize just the abusers.

But ISPs may have to consider offering anti-spam filters, anti-virus and IDS/IPS as part of the service.  Some already have this as a value added service or simply to protect their own network, but it may become critical in a usage-based service.  And, of course, spam filters, AV and IDS/IPS are significantly more intrusive - at least from the privacy perspective - than are bandwidth management appliances, even if the BMA’s apply DPI in a more brute-force manner (as one earlier post put it.)  BMA’s don’t look into the content for the purpose of surveillance and certainly not to store it for later retrieval as some blog posts try to lead people to believe (could you imagine the storage capacity you would need?  Only Google could do it.)  So, the Net Neutrality purists who preach the privacy side of the argument will have another dilemma to face.  And round and round we go.

Ultimately, until the day comes when we all have fiber to the home and bandwidth really is “free”, we will need to deploy a variety of tools in service provider networks to ensure service quality and reasonable prices.  Whether or not those tools are “fair” depends on what perspective you take.

You make it sound like it’s a choice between metered access and unlimited access. As I’ve pointed out elsewhere, the bulk of broadband services in Australia are neither metered nor unlimited, but rather limited on a gigabytes-per-month basis, and throttled back to dial-up speeds for the remainder of the period if the quota is exceeded. The price per month is generally fixed, but a higher price is paid for a higher bandwidth cap. They generally include spam-filtered email in the service, of course.

If you want to learn how decent broadband services can be provided without charging by the byte or sabotaging particular applications, look into the Australian broadband market. I don’t say that it’s perfect, but it’s clear that the US market can learn a thing or two from it—assuming that you’re looking for answers, rather than excuses, of course.

I use the terms “metered” or “usage-based” liberally to sum up anything that is not flat-rate no-usage-maximum “all you can eat” service, rather than try to come up with terms for th e 50 different permutations of, well, “usage-based” service.  And I have been a part or at least have seen many pricing models and technical strategies for Internet service deployed globally, not just in the US.

The language you’ve used assumes variable pricing for limited services: specifically, you introduce the “bigger bills” bugbear, and all the associated fuss over whether traffic was solicited or not. For plans which are fixed in price but bandwidth-capped, these are not noteworthy issues, and a light user can potentially save money relative to an “unlimited” plan. Your summary of usage-based plans seems to have treated the whole as though it were a particular subset: the subset of variably-priced plans. That subset is not representative of the whole.

Let me see if I understand you correctly. The impression that you’ve been giving me here is that a single one-size-fits-all unlimited plan is better than any tiered, usage-based system, including usage-based variants such as fixed-price plans that engage a throttle when a quota is exceeded. After all, you have predicted bad consequences if the FCC ruling results in a shift to usage-based services. Correct me if I misunderstand you.

If this is a fair description of your stance, then I disagree with it. Speaking for myself, I’m glad that I have a choice of fixed-cost usage-based plans rather than a one-size-fits-all unlimited plan. More objectively, the passing on of some bandwidth costs to the user in this way encourages moderation: a person on an unlimited plan has a perverse incentive to maximise his value by transferring data gratuitously, whereas a person on a tiered, limited plan has an incentive to pick a reasonable tier and ration usage accordingly.