baacamp07 topic: Internet Peering in New Zealand

Many (or, at least one) have already commented on David Cunliffe's sudden decision to review peering as an important issue to the New Zealand internet market. We all tried not to say, "well, duh". But we're geeks, of course we knew that was the real issue all along.

I think the problem is that the big players can basically cash in on their market dominance, then when any other player gets big enough to get there, they look around and think, "well, now we're up here, look at all these customers being reamed all around us... they'll thank us if we only ream them a little softer".

How do they keep their dominance? Well, one way is by interfering with the natural process of peering, effectively charging content providers to deliver traffic to and from their customers (who are also paying for the traffic). I think that there is an easy way to get around all of this, that represents an enshrining of what should be best practice and just good common sense.

What they should be doing is breaking up their netblocks into geographical segments. Refusing to route between netblocks is fine; why should I send this data from Wellington to Auckland for you? But refusing to accept traffic that is right at your door really has no excuse.

Tenet 1. aren't the edges of the network supposed to be the expensive bits?

So, you want to charge to provide me with data? I thought you already had a good network? Oh, you can't figure out how to deliver data without it costing the earth. What a shame you've slipped to being a lower tier provider. Well, tell you what, I'll build a network to you, and provide your network with service to all of the netblocks of my customers. Oh, would you like to borrow my uplink? I hear from your customers that yours must be poor.

So, anyway, the legislation corresponding to this would look something like:

6.3.1.2 Non-refusal of information exchange

    If a service provider gets a connection of a suitably prevalent network
    interlink type to the front door of the registered peering location of
    the internet netblock, the routing must not be refused, unless:

      A) another netblock which can route the traffic to the block is
         nominated, or advertised using standard Internet route discovery
         protocols such as BGP.

      B) the netblock peering location is in the process of changing, and the
         new peering location is to be used for new connections.

      C) another address is nominated for data exchange.

      D) another suitable address can be agreed with the exchanging party

    Peering addresses must not change more frequently than once every three
    months and never more than twice in a single calendar year.

Otherwise, it's like NZ Post setting up an address for service in Australia, and sending return mail to you Cash On Delivery.

I made up the three month / twice a year thing - somebody with more of a practical clue should pick good figures there. You could side-step this issue somewhat by saying they have to be approved peering spots, but that seems a little communist.

Tenet 2. if you're not peering in New Zealand, it ain't internet you're selling

How can you sell an "internet" connection if the vision that underpinned the founding of the internet is not just flaunted, but poked with a sharp stick and then squirted with juice from a bent lemon peel?

7.4.1.43 Internet services sold in New Zealand must have peering points in New Zealand

    If any service is sold in New Zealand which is advertised as connecting to
    the Internet, then the netblock must be one which has a registered peering
    address in New Zealand, unless another netblock is nominated from which
    routing to the block is allowed.

Anyway, those two tenets are basically it - and under the provisions of the existing Telecommunications Services acts and Roading acts, this should be all that is necessary - as any registered Telco should be capable (in theory, anyway) of running a cable to any registered address in New Zealand.

Hopefully what will then happen is that people like Telecom will break up their big netblocks into smaller ones as described above. The market for competition on transit will be opened wide. And it will be viable for those people who have the $900m to invest in deep sea cables to lay them. And we'll have our world-class high speed resilient internet backbone back.

Updates

I got a lot of feedback about these ideas from John Houlker from NZTE.

This proposal does not mean that content providers get a free ride

One of the problems with saying "free" is that "there is an old accounting rule - if something says free, rape it". Just like interest free student loans?

Well, anyway, you still need to pay someone to get the data out - we're not mandating hot potato routing (doesn't the argument just gain so much strength with all of these jargon terms?), but we are saying that you can't set the price. The market should set the price.

Because I'm sure there must be a flip side to that old accounting saying - "if you've got the market captive, charge what the market will bear"

Which set of netblocks? RIPE? any old /24?

For a start, what we're saying is that whoever is deciding where the netblock boundaries are (eg, RIPE), get to decide the smallest possible neutral peering partition. I think this is a good starting point. However routing within New Zealand can divide the netblocks up further - so, potentially, you could use allow a New Zealand registry to register these sub-divisions.

Here is where the tuning comes into it - Telecom could say they want "50 peering sites, just in Wellington" if they felt like making each exchange a peering point. It doesn't have to (and really, shouldn't) mean that Telecom is left with an unreasonable burden of forwarding. But equally they should not just say every /28 on their network is its own netblock.

Update: Telecom have identified 29 national peering points for people who want to play this game. Good for them!