Log in

No account? Create an account

Previous Entry | Next Entry

Interesting and extensive Ars Technica article about the emergence and details of large-scale deep packet inspection/packet filtering equipment, and their implications for net neutrality. I found it well worth a read.

In general, I'm a net neutrality supporter, because I see the Net as something on the level of a public utility, like the telephone, electricity, water/sewage, or even highway networks. However, the provenance of the Net as a stitched-together set of private networks (as opposed to the other examples which (to the best of my knowledge), when they became pervasive, were largely managed either by government-owned utilities or government-endorsed monopolies), offers a unique set of challenges.

A case can be made that certain applications (VoIP, streaming media, and the like) are time-sensitive and "deserve" higher priority, or that ISPs and backbone operators have a right to decide what their particular series of tubes will carry. However, given the way networking has evolved, the ability to filter, shape, or prioritise content by any given layer-7 protocol, especially as described in the article, inevitably entails the ability to do the same by destination, and - less certainly possible given the potential of encryption, but certainly appealing in some quarters - by content. The benign applications of such filtering left unchecked cannot, in my view, justify the inevitable abuses that would follow.

Since the Internet is not, however ubiquitous it has become, universally managed as a public utility, it is unarguable that significant numbers of ISPs and backbone operators have the profit motive more in mind than the health of the network. Further, the barriers to entry in this field restrict the ability of the free market to act as a corrective force - there is little in the way of alternate choice if you don't like the way your ISP chooses to restrict the use of its bandwidth, and even less choice if some upstream provider does the same.

It is thus, however paradoxically, necessary to employ government regulation to ensure freedom and preserve the Internet as we know it.



( Walk among 6 shadows — Cast a shadow )
Jul. 27th, 2007 06:51 am (UTC)
Myself, I think that the Net should be managed much like you put it, as a public utility. Basically just have the bandwidth like electricity currently is, it just flows to your home and you are free to do whatever you want with it, but with the catch that you pay for as much of it as you use.

If you want to stop the "exaflood", simply start charging people for their bandwidth on a per byte basis. If it costs a cent per megabyte, people will eventually catch on that, "Hey, this is starting to cost me a lot of money". It also gives a massive discount to those who don't use the internet as much.

Prioritizing content on a "layer-7" approach is never going to work. All I have to do to beat that, is just encrypt all of the data that I send out. Take Bittorrent for example, if every day they push out a new encryption/decryption key to the users of the service, then you will never be able to analysis those packets as they will look totally different than from the day before. (course in Bittorrent's case, you just have to look for lotsa packets talking between the same large group of users)

Now, you might argue that you can get around that by just throttling down any traffic that you can't identify, but, that has a whole lot of problems associated with it. As you will never be able to identify all programs accurately. New programs are created every day. And people like the government might not like you intercepting and decoding their top secret communications.

Now, to some of the specific protocols mentioned here:
Email - Hardly critical if your message shows up 2 seconds later than it otherwise would have, users will never know, so there's little point in prioritizing it.

VOIP - VOIP does have to be constant and realtime for it to work effectively, but, consider that a 56k modem is easily, easily capable of carrying a voice conversation over it (it used the same phone lines after all). So, this is actually very little bandwidth. The telephone companies aren't having any issues here, and their networks are ancient.

Streaming Media - The whole streaming media concept is very broken in my mind. It should all be done as a full download to the client's system, at which point they can view the media lag-free, whenever they want, and be able to jump around/rewind/pause without screwing things up or having to redownload pieces. Once it turns into a full download approach, then the real-time aspect goes away. This is where bandwidth should be throttled to make room for mission critical applications.

Gaming - Gaming does need real-time responsive speeds, but, it really doesn't actually transmit all that much data. This would be really hard to classify as each game is going to be unique.

As for the freedom aspects, I don't like the idea of the government or anyone else for that matter, being able to monitor my internet usage and communications. But, even without DPI, I can still intercept all of your packets, so, nothing really changes here, other than making the sorting a bit easier. Besides, they really aren't interested in most people's email anyways.

As for backbone operators being able to black-mail people like Google into paying more, that's just plain nasty business practices and ought to be illegal (price discrimination). There's no reason Google should be charged more or less than I am.
Jul. 30th, 2007 05:14 am (UTC)
I'm curious: What is a mission-critical application of the Internet, in your view? Which layer-7 packets absolutely must be moved as quickly as possible, and if that breaks someone's YouTube stream, so be it? The problem with prioritising anything is - who defines priority?
Jul. 31st, 2007 05:50 am (UTC)
Hmm, mission critical applications of the internet. It certainly wouldn't be the things that the vast majority of users use it for. Email, bittorrent, website-surfing, file-sharing, instant-messenging, etc, all of those can be delayed or slowed down without anyone caring or likely even noticing. The only things that can't really be interrupted are things like voip, gaming, streaming-media, none of which are all that critical if they suddenly stopped working. Though, with people switching from regular phone lines to voip service, that's becoming increasingly prevalent and thus important.

So, for mission critical things that use the internet, that would cause potentially bad things to happen if they screwed up:
1. Emergency voip calls. Easily filtered by prioritizing calls to 911 and such, since those could potentially kill people by delaying/dropping. Overall bandwidth is small, but, you want to ensure that the connection is made and maintained.
2. Things like online banking, interac transactions, and other banking stuff. Since if this went wrong, the economy would be paralyzed and there would be a lot of potential for harm.
3. Anything the government has out there that is needed to keep the country running.
4. Things like remote monitoring/control of the power grid, industry, and other utilities, that would be quite bad to go down.

Really though, the majority of the stuff on the internet couldn't be considered mission critical by any standard.
Aug. 2nd, 2007 03:40 pm (UTC)
It's a shame that ISPs are so mega-conglomerated and generally evil. It would be nice if it could be as simple as flocking to the provider that promised neutral tubes. Still wouldn't help with the backbones though.
Aug. 2nd, 2007 11:43 pm (UTC)
Yeah, that's the thing, while you can legislate choice in ISPs by opening up infrastructure (like was done to the telephone network), you still have the backbones to be dealt with, and you need lots and lots of cash to create/operate a backbone network, which limits entry quite thoroughly.
Aug. 3rd, 2007 01:46 pm (UTC)
Hmm, I guess even Vinge-style emergent mesh networks only take you so far, and you'd still backbone to get across the oceans. *pouts*

Well, we'll just have to wait 'til we get IP-over-ansible.
( Walk among 6 shadows — Cast a shadow )