Sure, let's build the NBN with technology that's not proven at scale

Comment Internet Australia's new National Broadband Network (NBN) policy advocates using a technology yet to be proven to work at scale.

The organisation last week decided to update its policy from one I surmise as “never compromise: only fibre-to-the-premises will do” to advocating fibre-to-the-distribution-point (FTTDP) rollout as better than fibre to the node (FTTN).

FTTDP, the organisation argues, has the virtue of bringing fibre closer to the home. “From there,” says CEO Laurie Patton, “existing copper could, if necessary, still be used in the short-term. Alternatively, copper wires could be replaced, at any time now or in the future, with fibre into the building. Such an approach would provide a future-proofed network and avoid the need for a costly re-build in 10 to 15 years’ time when copper is simply no longer fit-for-purpose.”

There's a bundle of contradiction right there, because in one paragraph Internet Australia argues both for a rebuild of the last 20m of copper and against any network configuration that requires rebuild.

The there's the fact that FTTDP is largely unproven, with only a few minor deployments worldwide. Internet Australia, like many others, has argued that the NBN should be built sooner rather than later, often using the argument that every day Australia goes without broadband is a day on which the nation falls further behind trading partners. But the organisation appears now to be advocating that we pause again and figure out how to do widespread FTTDP.

What might someone who can't even get ADSL think when told that they've been moved off the FTTN footprint and onto a yet-to-be designed FTTDP plan?

The idea that re-starting the network with FTTDP is sensible is also at odds with the organisation's argument that “the overall costs of construction have come down as NBN has refined its operating practices and achieved economies of scale, irrespective of the technology chosen.” Because of the inevitable cost decreases, the organisation argues, Australia should just have gone full-speed ahead and built a fibre-to-the-premises (FTTP) network.

The decision to adopt the multi-technology-mix of FTTP, fibre-to-the-node and hybrid fibre-coax for terrestrial connections, the argument goes, means “At the same time as we’ve seen the technology change Internet speeds delivered to consumers in other countries have been steadily rising and causing Australia to fall behind quite dramatically.”

“We have slipped to 60th on global rankings from 30th just a few years ago, according to the widely-quoted ‘State of the Internet’ report from content delivery network Akamai.”

Would Internet Australia see the nation's league table place remain low while FTTDP is planned?

And why is the organisation happy to use Akamai data in its public pronouncements but, in correspondence with The Register, to question its usefulness?

The Register has been corresponding with Internet Australia's vice-chair Paul Brooks about the organisation's aims. One thing we discussed was this OECD spreadsheet describing actual download speeds around the world. I was interested in the OECD data because it shows that as of 2014 South Korea's national download speed average of 23.6Mbps puts it streets ahead of second-placed Japan's 14.6Mbps. The OECD used Akamai, M-Lab and Ookla for its tests, the latter clocking Korea at 50.67 Mbps and others in the low forties. M-Lab data is widely divergent.

The point of my inquiry was that if real world download experiences in two nations widely considered key competitors for Australia are making do with those speeds, why should we complain about an NBN capable of delivering 100Mbps?

Here's a part of Brooks' response to my query:

Akamai measures tend to measure low - they are based on real browsing traffic, which is good, but tend to be biased to the low side due to the nature of the objects retrieved from Akamai caches - generally very small, frequently accessed objects like arrow icon gifs and jpegs, which don't let a high speed link get up to full speed before the object is fully transferred. Where Akamai caches serve large files like video, they tend to be segmented variable-rate formats that converge to the visual play-out rate, regardless of the available link speed. So Akamai rarely actually exercises a high speed link significantly.

All of which looks to me like Internet Australia is happy to consider Akamai a respected authority when Akamai data supports its arguments. But if Akamai does not advance the cause of “more fibre”, its authority somehow vanishes.

Internet Australia also failed to answer The Register's questions about how FTTDP might be paid for. The organisation points out that the cost of FTTN and FTTDP builds are now quite close, sometimes to within $400 per premises. But more than three million premises are set to receive FTTN and a similar number will be connected to the NBN with hybrid fibre-coax. If the NBN is to use FTTDP whenever possible, let's pick the round figure of five million homes, multiply that by $400 and then wonder where the extra two billion dollars will come from.

Internet Australia also chose not to directly answer our question about an FTTDP build's impact on broadband costs for subscribers.

Internet Australia and others have also protested FTTN's electricity-hungry ways, on cost, environmental and network fragility grounds. FTTDP requires subscribers to power the distribution points. Internet Australia is so far silent on the environmental and financial consequences of FTTDP.

I've previously argued that in the best possible world, Australia would build its NBN with fibre-to-the-premises. I retain that position.

But I also despair that debate in the field is making that objective harder to reach.

Internet Australia isn't helping with emotive language, confusing arguments and errors: the group screamed blue murder about Australia being denied same-day access to Netflix's House of Cards, an amateurish clanger that should have been avoided as the streaming company's global simultaneous release polices are well-known.

Now Internet Australia is jumping on the FTTDP bandwagon, promoting a technology that's recently achieved new prominence. I say recently because it's been argued as an NBN element since this 2014 paper. Internet Australia (and, to be fair, The Register^*) hasn't led debate. It's fanned it after nbn started discussion of FTTDP. And now it is advocating the riskiest approach yet suggested for a project it already criticises for delivering too little, too late. ®

^* In our defence, I think we were out front on G.fast and DOCSIS 3.1