Friday, April 26, 2019

Google Plus was about community and collaboration and killing it damaged users

Anti-trust law keeps large companies from stifling competition. Could consumer-protection law keep companies from simply killing services that many people depend upon?

Google Plus chief architect Vic Cundotra "sold" the project to Google CEO Larry Page by convincing him that Facebook was an existential threat and, when Google Plus was launched, Facebook took it as an existential threat and responded accordingly.

Both were wrong, because, while they overlapped functionally, they were not direct competitors. Facebook serves a social feed of posts from family, friends, and people an algorithm identifies as being like you. Google Plus also offered a feed but was more about collaboration and support of communities of common interest.

J. C. R. Licklider, who was responsible for much of the research that led to interactive computing and the Internet, anticipated Google Plus Communities by over 40 years, writing that interactive communities would "consist of geographically separated members, sometimes grouped in small clusters and sometimes working individually. They will be communities not of common location, but of common interest."

I barely use Facebook but did use Google Plus. I didn't pay much attention to my Google Plus feed but used Google Plus Communities extensively. I have an interest in the Cuban Internet, so created a Google Plus Community on the topic and joined several other Cuba-related communities. I also joined communities on other topics I am interested in, like the Internet in developing nations and satellite Internet service.

I am also a teacher and created a Community each semester for my students to share material relevant to our class and study together. My students and I also used Google Plus Hangouts on Air for collaboration and coordination.

At the start, Google Plus was not just a social feed -- it was a collection of services that also included Photos, Hangouts, and Communities. Google separated Photos and Hangouts before they killed Google Plus and could have separated Communities as well, but they did not.

Would Google Communities have been viable as a stand-alone service? Yes. The closest competitor would have been Facebook Groups, but Groups lacks key features like post categories and the Communities user interface was far superior to that of Groups. Revenue sources could have included community-member data, ads, an optional membership fee, etc.

What about fake news, spam and toxic filter bubbles? It's easier to manage those things when the unit of scrutiny is a community rather than posts in a social feed. A community will have a creator and perhaps one or more moderators. They would be the first line of defense against inappropriate content within a community. The operator of the community platform would guard against communities that were intended to violate platform rules.

If they decided not to operate Communities as a separate service, Google could have offered the code and data assets for sale to others or at least put the code in the public domain. Evidentally it was easier for them just kill Communities along with the social feed, but doing so damaged me and other users. I don't know how many Communities there were or how many members they had, but killing Google Plus caused the loss of a large amount of social and monetary capital.

Over the years, Google has killed 137 services, 12 apps, and 12 hardware offerings, beginning with Google Deskbar after a three-year run in 2006. Anti-trust law keeps large companies from stifling competition. Could consumer-protection law keep companies from simply killing services that many people depend upon?

Such a law would not be all bad for Google. If constrained, they would still recoup some of the value of their investment and it would instill confidence in other companies that were thinking of offering products that depended upon them. If I were a developer, I would be reluctant to build a product that depended upon Google with its track record of killing 161 offerings in 13 years. Google Plus also had a symbiotic relationship with other Google services. For example, Google Plus drove some Blogger and YouTube traffic.

Elizabeth Warren and others have suggested breaking up large Internet companies. If we do so, let's not throw away all of the pieces.

Thursday, April 18, 2019

Open data leads to competition

When evaluating proposed mergers and breakups, control of data should be considered along with market impact.

In a previous post, I spoke of Amazon's use of customer and market data in restraint of trade, but they are not alone. For example, leaked internal documents show that plans to sell access to user data were discussed for years and received support from Facebook’s most senior executives. Facebook gave Amazon extended access to user data because Amazon was spending money on advertising and partnered with them on the launch of the Fire smartphone. In another case, Facebook discussed cutting off access to user data for a messaging app that had grown too popular and was viewed as a competitor.

As former FCC Chairman Tom Wheeler points out in a recent post, proprietary data is a source of market control and he cites two examples where opening data has led to competition. In the US, a law mandating open access to video content enabled satellite companies to compete with cable companies and in the UK, open access to customer banking data led about 200 organizations to offer new services in its first year.

Wheeler's position is elaborated in Unlocking Digital Competition, a report from the Digital Competition Expert Panel convened by the British Treasury Department. Their data-related recommended actions are:
  • Establishing data mobility and open standards between services: overcoming network effects which cause markets to tip by requiring systems to ‘talk’ to each other using open, standardised formats. This will mean consumers can port their data between networks, interact with users on other, similar networks, and smaller firms can plug their services into those of bigger ones. New business opportunities will open up that use, manage, and combine data made available. Consumers, in turn, will have new choices of digital services, with switching made much easier.
  • Securing access to non-personal and anonymised data: tackling the data barrier to entry for smaller and newer firms, while protecting privacy. The power of bulk data driving economies of scale and scope is a key reason new firms struggle to compete and bring innovative services to consumers. Overcoming this barrier will allow the digital economy to remain dynamic.
These are only two of the 20 recommended actions in the 140-page report. Those actions are grouped under six strategic recommendations for the government:
  • Sustain and promote effective competition in digital markets, by establishing a pro-competition digital markets unit, tasked with securing competition, innovation, and beneficial outcomes for consumers and businesses.
  • Take more frequent and firmer action to challenge mergers that could be detrimental to consumer welfare through reducing future levels of innovation and competition, supported by changes to legislation where necessary.
  • Update and effectively use tools against anti-competitive conduct to help them play their important role in protecting and promoting competition in the digital economy.
  • Continue to monitor how the use of machine learning algorithms and artificial intelligence evolves to ensure it does not lead to an anti-competitive activity or consumer detriment, in particular to vulnerable consumers.
  • Conduct a market study into the digital advertising market encompassing the entire value chain, using its investigatory powers to examine whether competition is working effectively and whether consumer harms are arising.
  • Engage internationally on the recommendations it chooses to adopt from this review, encouraging closer cross-border co-operation between competition authorities in sharing best practice and developing a common approach to issues across international digital markets.
The two open-data actions mentioned above fall under the first strategic recommendation of promoting competition, but control of data is involved in the others as well. When Amazon acquired Zappos and Whole Foods, they gained access to data on relatively affluent shoppers. Facebook's acquisition of Instagram and WhatsApp and Google's acquisition of Waze also yielded data in addition to eliminating competition. The machine learning recommendation involves training data. When evaluating proposed mergers and breakups, control of data should be considered along with market impact.

The last strategic recommendation -- international engagement -- recognizes the global nature of the Intenet. (Note that the Digital Competition Expert Panel was chaired by an American). Nations like China and the US have different goals with respect to competition, but democratic, capitalist nations should strive to adopt compatible institutions and policies. In the era of Brexit and MAGA, we need to work with other nations -- I'd rather end up with two Internets than fifty.

Thursday, April 11, 2019

Amazon's orbiting infrastructure

The invisible eye of the marketplace
I have been following satellite Internet service since the 1990s, but I was surprised when I learned last week that Amazon had filed an application for a 3,236-satellite constellation of low-earth orbit Internet service satellites -- Project Kuiper.

I shouldn't have been surprised -- Amazon was an infrastructure company from the start.

In his first post-IPO letter to shareholders in 1997, Jeff Bezos pointed out that their distribution center capacity grew from 50,000 to 285,000 square feet and said their goal remained "to continue to solidify and extend our brand and customer base. This requires sustained investment in systems and infrastructure to support outstanding customer convenience, selection, and service while we grow."

Today Amazon infrastructure is used internally and is offered as a service to others. Their distribution centers are now highly automated and they distribute a lot more than books. Amazon also offers Web, cloud storage, shipping and delivery services, credit cards, a voice application platform, an affiliate retailer program, satellite ground stations, automated retail stores, pickup locations, Whole Foods stores and other things I am probably overlooking. Bezos personally owns the Washington Post and the Blue Origin aerospace manufacturing and spaceflight services company. (Blue Origin has a contract to launch satellites for Telesat, a Project Kuiper competitor).

Bezos' preparation for Project Kuiper was hiding in plain sight with the reference to road building in the Blue Origen mission statement: "We're committed to building a road to space so our children can build the future" and it should have become more clear when Amazon added fully-managed satellite ground station service to its Web Service offering. Amazon says Project Kuiper "will provide low-latency, high-speed broadband connectivity to unserved and underserved communities around the world.” That may be true, but it is the tip of the iceberg -- like saying Amazon saying in 1994 that they would deliver low-cost books to homes.

During the industrial era, infrastructure companies like railroads and oil and steel companies grew as quickly as possible in order to achieve economies of scale and create barriers to entry and profit from usage fees and sales. In the information era, data is as important as fees and sales. Esther Dyson pointred that out in 1995, the year after Jeff Bezos founded Amazon, and it was reaffirmed recently when Softbank founder and CEO Masayoshi Son justified his billion dollar investment in OneWeb's satellite constellation, saying "whoever gets the most data wins."

Stacy Mitchel has researched the ways Amazon has applied Dyson's insight. Since many people go straight to Amazon rather than use a search engine when shopping for products, Amazon learns what people want, what they eventually buy and don't buy and how much they pay. They use that market knowledge to decide what to feature in search results, which products to brand or manufacture themselves, which companies to buy, etc. Their size and information facilitate optimal and, in some cases predatory, pricing. Mitchel cites book sales and their zapping of Zappos as examples of the latter and also shows ways in which Amazon has used government subsidy. (Imagine the price war between Jeff Bezos and Elon Musk).

Amazon harvests data from all of their infrastructure offerings. For example, Netflix uses AWS. They even learned a lot about local demographics, real estate prices, labor costs, etc. when they invited cities to apply to be the site of Amazon's future headquarters.

But, what's wrong with this? Amazon is efficient and has kept prices low and their customer service is terrific.

That sort of reasoning has dominated US anti-trust enforcement in recent years, but it is partial and short-sighted. Prices that are low enough to maintain rapid growth suit Amazon well now, but in the long run, competition and transparency fuel low prices, efficiency and the broad distribution of wealth and income.

Industrial era concentration of power resulted in anti-trust action in the early 20th Century, but those were simpler times. In the letter to shareholders mentioned above, Bezos also stated that "there are significant opportunities to better serve our customers overseas." How can we achieve competition between global Internet service providers like Amazon, SpaceX, OneWeb and Telesat that are outside the jurisdiction of a single nation?

There is no simple answer to that question. We can't put the genie back in the lamp, but we have seen some success with government operation of neutral, free or wholesale infrastructures like roads, sidewalks and municipal backbone networks and Google has had some success with fair, wholesale networking in Africa. Europe is beginning to look for ways to encourage competition, transparency, and privacy.

Update 4/12/2019

In yesterday's 2018 letter to shareholders, Jeff Bezos reported that 58% of Amazon.com retail sales are made by independent third-party sellers, emphasizing that they "helped independent sellers compete against our first-party business by investing in and offering them the very best selling tools we could imagine and build," i. e., infrastructure. He singles out Fulfillment by Amazon, the Prime membership program, Amazon Web Services, database tools, SageMaker for machine learning and Alexa, which is built into 150 different products.

He also points out that Amazon remains a small player -- "low single-digit percentage" -- in global retail largely because nearly 90% of retail remains offline, in brick and mortar stores. But they are working on infrastructure for them -- Amazon Go stores.

He has convinced me -- forget retail, Amazon is an infrastructure company.

Amazon's infrastructure yields revenue and data, and that data can be used in restraint of trade, but Amazon is not all bad. As Bezos points out, they create jobs and valuable services as do others who use their infrastructure. Elizabeth Warren and others are calling for changes to US tax and anti-trust laws, but we need to be careful not to throw out the baby with the bathwater. (Full disclosure -- I loved Mrs. Maisel, the Big Sick, and many more and bought a terrific pasta cooker at Amazon.com yesterday).

The issue is further complicated by the fact that this is not a US issue, it is global. Amazon (and others) are Global infrastructure companies and China is a global infrastructure country.

Update 1/13/2020

Amazon competes with and gathers data from retailers who host their stores with Amazon. A similar situation exists in satellite Internet service where both Telesat and OneWeb have signed launch contracts with Bezos' Blue Origen. Will SpaceX also launch satellites for competitors and what sort of scrutiny is necessary in these cases?

Wednesday, April 03, 2019

Are inter-satellite laser links a bug or a feature of ISP constellations?

I hope each of these companies has someone in charge of thinking about what might go wrong with a single, satellite-based network providing fast, low-cost links anywhere on the global Internet.

Inter-satellite laser links (source)
OneWeb, SpaceX, Telesat and Leosat all aspire to be global Internet service providers using constellations of low-Earth orbit (LEO) satellites. Their success will require still-unproven technological innovation, but there are also political stumbling blocks.

OneWeb has already encountered significant political problems in Russia. Russia launched OneWeb's first test satellites and has a billion dollar contract for 20 additional launches, but Russian security officials are lobbying against OneWeb's offering service on the grounds that it might facilitate spying. When Anatoly Zak, an expert on the Soviet space program, investigated that claim, he concluded that "With the launch of the OneWeb constellation, the Russian rocket industry stands to earn millions, but the Kremlin is terrified at the prospect of unhindered access to the Internet by its citizens."

As far as I know, OneWeb is still planning to offer service in Russia, but they have had to make financial and technical concessions. They agreed to become a minority partner in the company that will market their service in Russia and, significantly, they agreed to drop the inter-satellite laser links (ISLLs) from their constellation and pass all Russian traffic through ground stations in Russia.

Dropping ISLLs, which are still unproven for this application, will simplify the design of their satellites and save development time and cost. It will also reduce satellite complexity, size, and weight and save power, but there will be costs.

OneWeb throughput simulation (source)
They will need more ground stations if they offer global service without ISSLLs. OneWeb founder Greg Wyler says they will have more than 40 such gateways, each capable of “seeing” satellites up to 4,000 kilometers away. A team of MIT researchers ran a simulation of a 720-satellite OnWeb constellation and they estimate that 71 ground stations would be required to reach maximum throughput. (Anatoly Zak estimates that four to six gateways will be in Russia and speculates that hackers may be able to illegally connect to satellites in neighboring countries). Dropping ISLLs will also add latency, especially on long-distance links.

What about the other would-be global LEO projects?

Telesat will retain ISSLs, but accommodate countries on a country-by-country basis. Erwin Hudson, vice president of Telesat LEO, says we "have the flexibility in our network control system to route traffic all kinds of different ways. There are no rules that traffic has to go over the inter-satellite links."

SpaceX is going forward with ISSLs, and they are also aware of the political problems. During a recruiting talk at the opening of their Seattle satellite-design office four years ago, CEO Elon Musk said "I'm hopeful that we can structure agreements with various countries to allow communication but it is a country-by-country basis ... it's not gonna take longer than five years to do that and not all countries will agree at first ... that's fine."

Leosat is also prepared to build gateways to make accommodations. For example, CEO Mark Rigolle says they would be willing to build a gateway in China to accommodate the government; however, that would be inconsistent with their primary marketing focus of providing low-latency, secure, point-to-point links globally.

These companies are all thinking about concessions they need to make in order to operate in countries that want to surveil citizens and control their access to information, but what about guarding against aggression? I hope each of these companies has someone in charge of thinking about what might go wrong with a single, satellite-based network providing fast, low-cost links anywhere on the global Internet.

During the recruiting talk mentioned above, Elon Musk said "[the constellation design] is a really difficult technical problem to solve so that's why we need the smartest engineering talent around the world to solve the problem and, you know, to also make sure we don't create Skynet." At that time, I asked "Would global Internet service providers require unique regulation and, if so, what should it be and who has the power to do it?" and said I was "less worried about Musk creating SkyNet than creating Comcast on Steroids," but I was naive.

Follow these links for background on the four projects: SpaceX, OneWeb, Telesat and Leosat.

Update 1/13/2020

OneWeb hopes to establish three ground stations in China. They have signed a "framework agreement" for one in Sanya, a city in southern China, and hope to have that and two others approved. They are also seeking partners to market and support their service. They plan a total of 45 global ground stations.