The Intellectual Property Law Bulletin at USF is holding a symposium on Net Neutrality. Here are the details:

Net Neutrality refers to free access to the Internet without discrimination based upon content, how often a user accesses the Internet, or the type of services and programs used.

The University of San Francisco School of Law Intellectual Property Law Bulletin is sponsoring The Toll Roads: The Legal and Political Debate Over Network Neutrality, a symposium to increase awareness about network neutrality, bringing together lawyers, academics, economists, and technologists for a balanced debate on the issue. Panelists include Tim Wu, Richard Clarke, Lawrence Spiwak, and many others.

When: January 26th, 2008 8 AM - 7 PM
Where: Fromm Institute on the University of San Francisco main campus
Web: http://www.netneutrality2008.org
Cost: Professionals (6.0 Units MCLE Credit): $100
Non-professionals: Free - $75 (see registration page for details)
Register: http://www.netneutrality2008.org/Registration.html

My current intention is to attend. It should be interesting. I'm not a huge proponent of mandated neutrality, and I'm bothered that most advocates (and some of the statements on the symposium site) suffer from the seeming belief that the Internet is and has been neutral, whereas it isn't and never has been. Hopefully, someone will be trumpeting that side.

the seeming belief that the Internet is and has been neutral, whereas it isn't and never has been.

What do you mean by this?

Obviously, it's a bit tough to answer the question without defining net neutrality, and that's one whole panel on the discussion. However, what set me down this path was the following on the symposium's front page:

Should Internet Service Providers continue to drive the same “neutral” course, giving all search requests and other uses of the Net equal treatment? Absent legislation, will these ISP’s create Internet “toll roads,” giving preference to certain types of information? It is these types of questions that are at the heart of the debate on Net Neutrality.

I deleted what became a critique of this notion, that somehow this hasn't always (at least in modern Internet times) been done. All uses of the Internet have never been given equal treatment. Have you ever been able to get your transit provider to get multicast to you? Are you on the MBONE? Does your protocol use UDP, because it's generally deprioritized in favor of TCP. What if you want to run an autonomous network with multiple routes via BGP, but you can only secure a class C? No good, because: a) you're not going to get it; and b) no one is going to accept your routes. What if you've built your own backbone and you want super-fast interconnectivity? Think the big guys are going to peer with you unless you've got something to offer? Ten years ago, what if you ran a second-tier backbone and didn't want to interconnect with UUNet and BBN at a public peering point. Too bad, because the big guys weren't going to give you good private peering, so you, your customers, and your applications would lose packets in one direction or another.

And that's just the easy ones off the top of my head.

"Have you ever been able to get your transit provider to get multicast to you?"

Sure, if you pick the right provider it's generally not a huge problem... the list of multicast enabled providers in the US is small and shrinking, but... (verio/ntt-america) comes to mind.

Also: "Does your protocol use UDP, because it's generally deprioritized in favor of TCP"

generally deprioritized where exactly? Having worked at a largish ISP for a few years and knowing a bunch of other large ISP folks for years I've yet to find any significant portions of the 'net that actually 'deprioritized' UDP traffic. There are many, many services that rely solely on UDP (dns and most streaming media things, save the current wunderkind - flash-movies).

"What if you want to run an autonomous network with multiple routes via BGP, but you can only secure a class C? No good, because: a) you're not going to get it; and b) no one is going to accept your routes"

what?? If you have host justification, or a business need for multi-homing you can get a /24 from ARIN (RIPE or APNIC have similar policies) and an ASN from same... This happens every day, in many different places... A /24 (class-c) is the defacto standard longest route accepted across the board, certainly there are places that do NOT accept these (for memory/cost reasons) but they carry a default route (or miss out on lots of things on the internet...) so they aren't really part of the problem.

Your points about peering are mostly true, though the end bit ("...give you good private peering, so you, your customers, and your applications would lose packets in one direction or another"). What really happens is that there is first a business decision to peer or not (will each side essentially have a zero cost in the traffic exchange) then a port-cost decision.

Today most of the larger networks will only really peer with like-sized networks, everyone else is expected to purchase transit from someone and either insure that they get full routes, or carry a default.

On the topic of net-neutrality the things I'd point out are providers that use blacklists to filter email (only email that lands on their mail complexes since transit tcp/25 is never really messed with unless the source is special...)

anyway, I look forward to your thoughts and information about the conference/symposium, please do report back! (I had wanted to go to it as well but work things conspired against me)

This is what I get for popping off in comments. I should have been clearer on a few points. However, I'm still open to errors; you've been more in touch with some of these issues than I.

"Have you ever been able to get your transit provider to get multicast to you?"

Sure, if you pick the right provider it's generally not a huge problem... the list of multicast enabled providers in the US is small and shrinking, but... (verio/ntt-america) comes to mind.

Bear in mind that I was making the point that the claim that today's Internet is neutral -- which suggests to me to mean anyone who wants to hook up -- or that the Internet in recent years (let's say the last 15) has been neutral is flawed.

So, the fact that I can go get a provider who is multicast-enabled only seems to support my point. If I can't get multicast from any provider to any provider (even in CONUS), then there has been a decision made and all traffic is not treated neutrally.

Also: "Does your protocol use UDP, because it's generally deprioritized in favor of TCP"

generally deprioritized where exactly? Having worked at a largish ISP for a few years and knowing a bunch of other large ISP folks for years I've yet to find any significant portions of the 'net that actually 'deprioritized' UDP traffic. There are many, many services that rely solely on UDP (dns and most streaming media things, save the current wunderkind - flash-movies).

Yes. I'll chalk this one up to writing too fast for my own good, because what I really had in mind was router deprioritization of handling ICMP under high loads.

what?? If you have host justification, or a business need for multi-homing you can get a /24 from ARIN (RIPE or APNIC have similar policies) and an ASN from same... This happens every day, in many different places... A /24 (class-c) is the defacto standard longest route accepted across the board, certainly there are places that do NOT accept these (for memory/cost reasons) but they carry a default route (or miss out on lots of things on the internet...) so they aren't really part of the problem.

Well, a class C space and a /24 aren't really equivalents; can anyone really get a class C anymore? (I didn't mean to suggest that one couldn't get an ASN.) According to ARIN, they won't assign blocks longer than /22:

4.3.2.2 Multihomed Connection

For end-users who demonstrate an intent to announce the requested space in a multihomed fashion, the minimum block of IP address space assigned is a /22. If assignments smaller than a /22 are needed, multihomed end-users should contact their upstream providers. When prefixes are assigned which are longer than /20, they will be from a block reserved for that purpose.

I guess I'm surprised that people are taking /24 routes at all, especially if ARIN won't assign them. If that is the case, then that is a change from when I last worried about such things. However, it certainly is the case that only shorter routes were accepted in the past by major players (Sprint in the mid-90s comes to mind, though I believe your former employer was in the mix, too.) Are you sure this isn't only the done in cooperation with transit-supplied IPs or without special arrangements?

Today most of the larger networks will only really peer with like-sized networks, everyone else is expected to purchase transit from someone and either insure that they get full routes, or carry a default.

Right. Everyone's traffic is equal, but some are more equal than others.

Your points about peering are mostly true, though the end bit ("...give you good private peering, so you, your customers, and your applications would lose packets in one direction or another"). What really happens is that there is first a business decision to peer or not (will each side essentially have a zero cost in the traffic exchange) then a port-cost decision.

Of course, it's all a business decision. However, that business decision values traffic and treats some traffic differently than others, based on business decisions. That doesn't strike me as neutral.

Sorry to have posted and run...

I see what you're saying. Details aside, if everything you say is true then what we really seem to have is an issue of definition of terms. Because maybe you're right, and certain kinds of Internetworking is less equal to other Internetworking, but then you seem to be talking about discrimination on a different level than what the debate about "network neutrality" is about. Perhaps there's some overlap when it comes to discriminating based on the shape of the packet, but even then I'm not sure that discriminating over aspects such as TCP/UDP is the same thing as discriminating based on the application using a particular packet (e.g., P2P, HTTP, etc.) and it certainly isn't the same thing as discriminating based on the particular content carried by the packet (e.g., independent content v. Time Warner content...)