When we think about the problem of achieving affordable access to the Internet for all, the discussion often focuses on broadband targets.  These targets are moving goalposts as infrastructure improves.  Broadband used to be defined at 256Kbps, now it might be 2Mbps or 5Mbps or something else depending on who you talk to.  Nobody wants to be seen as providing second-class broadband to their country.  And yet, the higher we set our targets for broadband, the more ambitious those goals become.  This is particularly true when we are talking about rural regions with variable population distributions with low ability to pay.  I have come to wonder whether simple broadband targets are the right strategy.  Perhaps we need something more nuanced.

First I want to suggest that slow Internet is still extremely valuable Internet.  Only very modest speeds are required for the basic operation of enormously population messaging services like WhatsApp, WeChat, FB Messenger, et al.  The growth of those services in the last few years has been nothing short of phenomenal.  It is with that in mind that I along with Steve Esselaar and Christoph Stork have been pitching the value of making 2G Internet access available for free to anyone with a data-capable phone.  You can read more about that idea at Zero-Rating: A Modest Proposal or watch the 10 minute pitch I made to the Mozilla Equal Rating Challenge last week.

It is an interesting thought experiment to ask yourself which is the more equitable strategy: connecting 80% of the population at 5Mbps or 100% of the population at 2G speeds.  The answer of course is that it is not an either/or question but that both strategies might be important.

And of course, 2G, while essential and extremely valuable, is not enough.  That got me thinking about where the value really exists on the Internet.  And for me the other place of high value exists at either end of the broadband spectrum.  I value small amounts of bandwidth for texting, emailing, etc and I value video, more specifically I really like YouTube.  I don’t like singling out a particular platform but no one else is delivering value the way YouTube is at the moment.  And at the low-end of the value proposition somewhere are top-heavy news sites like CNN.  I don’t mean to pick on CNN, there are hundreds like them; sites that deliver 20KB of reading material while taking up 5MB of capacity.

A word about YouTube.  I think YouTube is one of the pedagogical marvels of the decade.  There is virtually nothing you can’t learn to do on YouTube whether it is learning to be a carpenter, to fix your car, to write music, to be a magician, to fix electronics, to dance, to understand particle physics, the list is endless.  Chris Andersen of TED understood this six years ago in his talk on how YouTube is driving innovation. But it has gotten even better since then. Nadiya Hussein, who won The Great British Bake-off in 2015 is credited with learning to cook via YouTube videos.  Examples like this are legion. In communities where no affordable broadband exists, people travel to areas with free or cheap WiFi just to download YouTube how-to videos to take back to their communities.  Some even take requests.  I see this as well in my sons who are between ages 11 and 12.  They acquire skills on YouTube that frankly amaze me and they put more effort into learning there than they do in school because they are tapping into what they are passionate to learn about.  YouTube is also valuable because it bypasses a lot of local language and literacy challenges by allowing people to express themselves directly.  It’s not a panacea in this regard but it is a real vehicle for local expression.  Again I want to emphasise that it is not YouTube in particular that I am advocating for but the value that YouTube is providing right now.  What it might be in the future is another thing entirely.  Platform diversity is just as important as content diversity.

So if I can’t have the whole Internet, or if I have to decide what I get first, I would choose slow and fast Internet. My graph of Internet value vs speed looks something like this one on the right.  If my value framework is true, does that tell us something different about how to prioritise the development of Internet infrastructure?  For my it places a clear emphasis on getting everyone at least 2G access to the Internet but what does it mean we should do at the other end?  Conventional wisdom says invest in more broadband networks, 3G, 4G, 5G and so on.  In the last year WiFi networks have come to challenge that orthodoxy in Sub-Saharan Africa but, with a few notable exceptions, they tend to thrive in urban areas where there is existing high-speed backhaul infrastructure.  Getting YouTube or Vimeo or other video platforms to sparsely populated rural areas in an affordable manner is a real challenge.

That is why I am so excited about a pilot project that Google have in the Philippines called Google Accelerator.  Yes, not the catchiest name in the world but still an amazing project.  Google Accelerator is what I would call a micro Content Distribution Network (CDN).  You’ve heard of Akamai, CloudFlare, Google Global Cache and other CDNs that work at the network operator level and provide national-level caching for petabytes of content.  A micro-CDN works at the small organisation level.  Imagine a TV set-top box that is WiFi-enabled.  Now picture it with a 4 terabyte drive that is packed full of YouTube videos.  Anyone who connects to the WiFi access point of the set-top box gets instant high-speed access to YouTube.  Suddenly the high-value end of the Internet is available anywhere.  Of course the videos need to be regularly updated and that is challenge.  The ingenious part of Google Accelerator is that the content is refreshed via DVB satellite.  Pretty cool.  Google Accelerate is also remarkable because the solution is very affordable.  The components are commodity devices that might cost a few hundred dollars in total.  The challenge lies more in the refresh of the devices and what that costs.

Google aren’t the only ones who get the value of micro-CDNs.  Kenyan startup BRCK also seem to have pivoted to a content-based strategy in which their latest WiFi device the SupaBRCK which also has a hard drive specifically for local caching of Internet content.  This doesn’t mean there aren’t challenges.  Platforms like YouTube and others can be particular about allowing anyone else to cache their content.  Coming up with a licensing strategy that allows for independent micro-CDNs may be necessary.  Also, the trend to HTTPS encryption for all websites is a factor.  HTTPS encryption is great for blocking casual surveillance by bad actors but presents challenges for CDNs that can’t see what content is being browsed in order to cache it.  Large CDNs like Cloudflare solve this by decrypting and re-encrypting traffic as it passes through them but it is likely not the simple for micro-CDNs.

So what do you think?  Is the Internet U-shaped in terms of value vs speed?  Does that suggest different ways of prioritising infrastructure development?  This article was written partly as a provocation to test this idea that has been growing on me for a while. I’d love to hear rebuttals, affirmations, and/or alternative theories.




Posted by Steve Song

@stevesong local telco policy activist. social entrepreneur. founder of @villagetelco
#africa #telecoms #opensource #privacy #wireless #spectrum #data

  • lpress
  • Tome

    Good idea.
    I agree that speed of 500 Kbps should be free for all, because it’s enough for wiki and some other small consuming sites.
    Those which would use that speed should have unlimited internet (offcourse).

    For watching YT it should be higher, but this would be good step till it become affordable to have high speeds (even 2Mbps would be enough if it’s unlimited traffic for 30 days) or when wifi will be available everywhere for (almost) free.

    In Croatia before 5 years there were rare cafe bars, hotels with free wifi, but now is almost in all.
    In my town you don’t have free wifi on squares (was but turned off and not sure will come again) or similar, but have in one government building free internet cafe, so computers and internet connection. (talking about places which someone from outside my town/region can visit for free net).
    Still mobile internet isn’t that cheap here, but it goes every year cheaper.

    And after few years we finally got affordable internet with speeds higher than 10 Mbps.
    Still there is need to build fibre network to smaller villages and right now government and private player (Deutche Telecome) are in battle who should build it (it will be built from EU fund).
    T-HT (DT) has almost monopoly on fixed internet, but it’s good that it goes down.
    Our government sold T-HT and all infrastructure to DT and that was mistake.

    Here it goes slow, because here they have some ”better” things to do. :/

    I hope that in future there will be some xy number of operators which will operate in all countries in world(or in EU asap), so on that way they could have enough income to invest in poorer countries and have same QoS in all countries.
    As we can see some left Africa market (Airtel) and some come (Vittel). (pretty odd that Chinese telecom operators didn’t go outside China)
    This goes step by step in EU and it should go faster.
    Outsource towers and make all MNOs to MVNOs and grant EU license for all EU states, so those which now are on EU market get license to operate on other EU markets where they are not present.
    There will be some consolidation for sure, so on this way it could come sooner.

    Greetings from Croatia.

  • Tome

    News about fibre Uganda and South Africa/Botswana

    South Africa/Botswana
    Lastly, South African state-owned operator Broadband Infraco
    has revealed that its fibre-optic backbone network currently spans
    14,661km, with maximum capacity on some of the routes going up to
    170Gbps. The operator also said that it has completed the upgrade of the
    fibre link from Oberholzer to Ramatlabama, thus allowing more services
    to be provided to neighbouring Botswana.


    The National Information Technology Authority of Uganda (NITA-U) has announced that Phase Three of the National Data Transmission Backbone Infrastructure and eGovernment Infrastructure (NBI/EGI)
    project – comprising the deployment of 756km of fibre-optic cabling –
    was completed in December 2016, covering the districts of Masaka,
    Mbarara and Kabale, and extending connectivity to neighbouring Tanzania
    (at Mutukula) and Rwanda (Katuna). The next stage of the NBI
    project, aiming to link Atiak with Karuma via Moro, Arua and Pakwatch,
    is scheduled to be completed in the 2017-18 financial year, while Phase
    Five (Gulu-Kitgum-Kotido-Moroto-Napianyenya-Siroko-Mbale,
    Siroko-Kapchorwa and Soroti-Moroto) is currently in planning stage.
    ‘With the completion of Phase Three of the NBI
    there is development of an alternative geographically redundant route
    to the undersea cables connecting to the global communications
    infrastructure. There is also anticipated reduction in the costs of
    internet bandwidth as a result of increased competitiveness through the
    increase in the number of upstream providers available to IT providers
    in Uganda,’ the regulator said in a press release. TeleGeography notes
    that the government contracted Huawei for the deployment of the backbone network in late 2006; the NBI/EGI
    project was due to take 27 months and cost UGX201 billion (USD55
    million) to compete, though it subsequently experienced a number of

  • Man, there is a lot going on in Uganda these days for Fibre. NITA’s announcement, plus UETCL’s and Facebook’s tie-up with Airtel and BCS

  • Tome

    I read about many new fibre things in Uganda.
    And Google project too.

    Hope that it will decrease price of internet in Uganda and other neighboring nations (South Sudan, DR Congo, Rwanda).

    We will see.

    Is there any plan to update afterfibre map?

  • There is indeed and soon. I generally hang on though until I can convince someone to share an actual map of the planned network. I don’t know why operators are so reluctant to share fibre maps. It is the best possible kind of advertising for them.

  • Tome


    Finished submarine cable from Algeria to Spain.

    Construction work on the 560km Oran-Valencia (Orval)
    submarine cable linking Oran in Algeria to Valencia (Spain) was
    completed on 31 March, daily newspaper Levante El Mercantil Valenciano
    writes. The system was initially scheduled to be RFS
    in October 2016, but was subsequently delayed due to ‘bureaucratic
    burdens’. As previously reported by TeleGeography’s Cable Compendium,
    the Ministry of Posts, IT and Communications (MPTIC) and equipment vendor Alcatel-Lucent (now part of Nokia)
    signed a turnkey agreement for the construction of the Orval
    fibre-optic network in May 2015. The 100Gbps system will deliver an
    ultimate design capacity of 20Tbps when it enters commercial services.

    Is it easy for you to change some part, on your afterfibre map, of current cable from red to yellow?
    F.e. That one part in Guinea which finished last Autumn or winter? Didn’t read that it’s active, but should be if it’s finished.

  • Yes, very easy to toggle the active/planned status of the cables. I originally had that network in Guinea marked as live but on the feedback from some folk working there I set it back to planned as they were convinced it wasn’t live. You wouldn’t think it was that hard to establish. Will look into it.

  • According to the FCC, 99.1% of US census tracts have 10 Mbps service even if we don’t count the 12-15 Mbps satellite services.

  • High capacity fiber is for backhaul and interconnects, it’s not residential service. The US installs 19 million miles of fiber every year according to the CRU group.

  • Thanks for sharing that Richard. The focus of the article, which I obviously didn’t make adequately clear, is not on the industrialised world.

  • I love this post, Steve.

    I’ve been a quiet fanboy of the 2G-as-universal-service since you’ve all started pitching it. From my perch, I see a *ton* of impact occurring at the straightforward 2G connectivity level that often get overshadowed by the hype cycle story of the day—there are 100 stories of basic messaging success stories for every 1 of real-time video conferencing (not to take anything away from conferencing!)

    One question/nuance — is it that connectivity value is U-shaped, or is it simply plateaued at various steps. You could reason then that where you’re at in a particular step certainly *feels* like a decline in value, but it’s really only a relative decline to where you’re at on the particular step, not an absolute one. So being close to the edge of the next step is particularly frustrating because you can see how close you are to the next step, but it’s not an absolute value subtraction from the step just below you (you can feel this tension in action when you’re on a tenuous connection with high latency trying to watch the lowest bitrate YouTube video as possible!)

    This isn’t a refutation of the U-shaped Internet, just a nuance of it. The model has a lot of explanatory power — thanks!

  • Thanks Gabriel. You might be right. A step change graph would be a better way of expressing it. I certainly didn’t mean to imply that value decreases but just that it doesn’t linearly increase. A U-shaped graph would express rate of change of improvement. Thanks for that feedback!

  • David Johnson

    Great post Steve. Something which I’ve been exploring for a while now is also that value of upload caching due to locality of interest. So if somebody shares a video – there is a very high chance that somebody in the higher speed intranet on the campus/neighbourhood/village will want to access this – there are interesting studies that show how we are 10x more likely to make a voice/text/data connection to somebody 1 km away compared to 10km away and 10^5 times more likley compared to somebody 1000 km away. So if you upload a youtube/facebook video there is massive value in caching in locally. We built a primitive version of this in 2012 – I have a PhD student who is now looking at a much more robust version of this for all manner of secure content disttribution (content provider local village cloudlet, village cloudlet village cloudlet, village user village cloudlet etc.) in areas with poor/expensive Internet gateways. I wonder if Google would extend their Google accelerator concept to upload caching.

  • Hi David. I would be very interested if you have pointers to any of that research on proximity. I have just heard that Google have abandoned the Accelerator project or at least ceased moving forward with it. On the bright side, this does open up an opportunity in the market 🙂

  • David Johnson

    This is the the paper on proximity with the 1,10,1000km probabilities of a forming a social tie: http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0016939 – sad to hear about the acclerator project – but this is indeed even more motivation for us to press on 🙂

  • Thanks David!