The FT could not be more wrong about Brazil and the Internet.

The FT is worried about the Internet, and specifically what the Brazilians are up to with it as a result of the Snowden disclosures.

I am not. Details of Brazil’s very successful policy with regard to the development of the Internet are here, in a fine post on the Renesys blog. Basically, we’re seeing the conflation of two things here – the US’s genuine concern for the Internet, and its equally genuine concern for the interests of Silicon Valley, Hollywood, and the spooks.

What is the Internet? You can always start a good row on the NANOG list by asking for a definition. But I don’t think there is much real dispute that it’s fundamentally a set of interconnected autonomous networks. The key protocols overlay diverse media bearers and underlay diverse applications, while the routing system unites Autonomous System numbers, defined as networks having an independent routing policy. It differs from, say, the mobile carriers’ GRX and roaming hub infrastructure, or the older international voice interconnection systems, in that there is no central intermediary or necessary hierarchy. If AS X and AS Y want to interconnect directly, they can do so as long as they can physically reach each other.

As such, more local interconnection is always and everywhere a good thing for the Internet. Back in the 90s, although the rhetoric of radical networking was at its peak, the system was in fact heavily dependent on central intermediaries, the so-called Tier 1 operators. As a very rough generalisation, you could say that the Internet worked like it was meant to among US universities, government users, and research centres, and everyone else was eventually a customer. This had distinct geographical and political consequences – supposedly, the topological centre of the African Internet was the 111 8th Street carrier hotel in New York City, or the LINX in London, depending on who tells the tale. The convenience of this for the NSA should be more than obvious.

Those days are long gone. The Tier-1 operators no longer rule the earth like they once did, and anyway the club of Tier-1s has a very different membership now, with major players including Tata of India, Telecom Italia-Sparkle, and PCCW of Hong Kong where once it was a nearly total American monopoly. What changed? More competition, for a start, but also much more interconnection between customer networks. I mentioned the LINX, the London Internet Exchange. IXen are membership organisations that provide for peering between their members, and their spread has been a major factor in the quantitative growth and qualitative development of the Internet worldwide. Another important issue is the increasing tendency of eyeball networks – consumer ISPs – and content networks to peer directly.

To understand why this is a good thing, I recommend another Renesys post. Shorter paths equal lower latency, and more diverse paths equal greater resilience to disruption.

Latin America was rather late to this development. It exhibits the traits of the 1990s Internet to an extreme degree, with both nearly all its traffic to and from the wider world and enormous amounts of intra-regional traffic passing via Terremark’s NAP of the Americas in Miami. The politics of this, again, ought to be obvious, especially from a Latin point of view informed by their distinctive intellectual tradition. The economics and engineering of it are not good either. Locating hosting, applications, or content delivery infrastructure in a country that reaches its neighbours via a 5,000 mile submarine cable round trip is asking for trouble, and a structural barrier to scaling up an Internet company anywhere in the region.

Hence Brazil’s effort to create local IXen and ISPs and to build more regional fibre links, which has been a great success (check out the numbers in the post!). Not only do they want better connectivity, they also want participation in the Internet rather than just consumption of Google searches and lolcats. A good measure of this is AS numbers per capita. Theirs are rocketing.


(If you find this interesting, a more detailed technical report on it has just appeared here)

So, to summarise so far: there is nothing natural, open, or free about sending traffic from Sao Paulo to Rio via Miami. And the Internet founders did not intend and do not want this. It is bad economics and bad engineering, and the rest of the world left it behind years ago. Brazil is right to do everything possible to get rid of it, and its efforts are already bearing fruit. I think they are also right to identify it as a form of underdevelopment imposed on them by the rich. IETF and ISOC are very clear that peering and local interconnection are nothing but good.

We have an excellent counter-example to Brazil in the Renesys data set – Mexico, which had until very recently a private monopoly of telecoms, no regulator, and no IX. Mexico has far fewer AS numbers per capita and far less growth in this metric.

The FT is confusing the entirely healthy development of the Internet as it was intended to work, and indeed does in Europe, with the Great Firewall of China or the lesser firewalls of Saudi Arabia, the UAE, or Iran. In doing so, it is letting the US get away with using the rhetoric of free speech to maintain an exorbitant commercial privilege, and of course to tap Dilma Rousseff’s phone whenever this seems expedient. This agenda permits freedom in a negative sense, but at a real economic price and at the cost of giving up participation and control.

And it serves, probably unintentionally, another agenda – the Chinese or Iranian approach, in which free speech on the Internet is a political threat that must be censored and an open door for US competitors that must be closed by censorship in its secondary function as a non-tariff barrier. This one can work economically – note the huge Chinese Web 2.0 companies like QQ and Sina Weibo, or the roaring growth in AS numbers within Iran – but it is deeply illiberal and anti-democratic.

There is an alternative to this unattractive pair of options. That is real Internet development, like the Brazilians are doing, like Kenya is doing, or like Europe did in the 2000s. It must not be demonised by conflation with Chinese opinion management.

3 thoughts on “The FT could not be more wrong about Brazil and the Internet.

  1. Pingback: Links Thanksgiving 2013 « naked capitalism

  2. Pingback: Admin note: Latin American Internet | A Fistful Of Euros

  3. We have some more critical comment from “Trevisa”:

    People outside of Brazil are so wrong when they think brazilian government help in any way to the development of internet infrastructure. The maximum they did so far was promoting an initiative to finance expansion of fiber networks, but this is a project that is only reachable by big national players. NET has approved more than a hundred projects to build and operate its own CLOSED fiber and HFC infrastructure.
    There is ZERO incentive to small ISP´s.The rocket high growth of local ASN numbers are simply explained due to 2 factors:

    1 – You need a license to operate telecommunications service in Brazil. It´s an SCM license issued by Anatel. It used to cost 9 thousand reais, and now it cost 400. This was a major incentive to the small ISP´s to formalize and legalize themselves.
    2 – After reaching formality by being legally certified, it´s natural for them to become more professional at their day to day operations, and this means having their own IP block, which leads to a mandatory issuing of an ASN.

    In Brazil there are more than 4 thousand SCM licenses issued, and an estimate 12 thousand total ISP´s, or 8 thousand “illegal” ISP´s that uses someone else IP range.
    Considering that every ISP should have their own ASN, you can see a huge growth potential.

    Many ISP´s are, independent of government´s efforts, laying their own fiber network. Actually, against all odds they are doing this, because in every city in Brazil (more than 5400 cities exists here) there is a different regulation for using poles or undeground space. It´s HUGELY bureaucratic to be within the law and even worse to get a financial subsidize to build the networks.

    After coming across the last mile barrier, you have another fight when you try to buy IP Transit. Except for the capital of Sao Paulo the city of Sao PAulo and few other capital cities and surrounding areas, it´s extremely expensive to buy IP connectivity be it Transit or Transport. Some ISP´s are growing to a size that allow them to buy connectivity from their original location, be it in Amazonas, Ceara, or Rio Grande do Sul to the nearest IX and to the Sao Paulo Internet Exchange Point, the largest in Brazil.
    Sao Paulo´s IX is so large that it´s more than 7 times larger than all other brazilian IXen combined. It has grown from 100 Gb peak at the beggining of 2013 to 400 Gb peak at the of beginning of 2014. Think about it. It´s simply crazy to have every ISP in Brazil to interconnect in Sao Paulo. This reflects the lack of infrastructure connecting different regions and states and the strong hold the Tier 1 Operators have. Brazil is stuck in the past besides the attempts from few ISP´s to brake this barrier