By Lambert Strether of Corrente.
Democracy Now has an interesting interview with Clay Johnson, co-founder of Blue State Digital, the firm that built and managed Obama’s online 2008 campaign for the presidency in 2008, and then director of Sunlight Labs at the Sunlight Foundation. Johnson has been all over the Exchange debacle coverage (see here, here, and here). I’m going to take a look at some questions raised by Johnson’s very layperson friendly radio discussion.
But first, I should caveat that Johnson — as a subject matter expert in government procurement — was very helpful to me when I was trying to dope out ObamaCare’s CMS contracts last summer. (Alas, as matters turned out, it was dig into that story or — the life of a big-time blogger! — paint the house, so I had to paint the house.) And although in a very much past life I’ve done government contracting, I’m very much a content person, not a programmer. So, two strikes against me for taking on this topic!
So that said, let me start with something Johnson said in the DN interview that absolutely makes sense to me:
But when government is building software like this, it ought to be built out in the open. It ought to be built with a licensing system called “open source,”[*] so that the public truly owns it. You know, if my tax dollars are going to something, then it ought to belong—if the public’s tax dollars are going to something, then it ought to belong to the public.
This seems utterly unexceptional to me (modulo, off the top of my head, grandfathered Commericial Off-The-Shelf Software, and modulo a discussion about software that affects so-called “national security). Public code for public purpose, as it were. However, note that Johnson makes an ethical claim (“ought… ought… ought… ought”) and hence a policy and ultimately a political claim, and not a technical claim (which will become important in a moment).
So how would that happen? Johnson writes in a Times Op-Ed:
The president should use the power of the White House to end all large information technology purchases, and instead give his administration’s accomplished technologists the ability to work with agencies to make the right decisions, increase adoption of modern, incremental software development practices, like a popular one called Agile, already used in the private sector, and work with the Small Business Administration and the General Services Administration to make it easy for small businesses** to contract with the government.
Large federal information technology purchases have to end. Any methodology with a 94 percent chance of failure or delay, which costs taxpayers billions of dollars, doesn’t belong in a 21st-century government.
This again is an appealing vision. In fact, if it were easy for small business to contract with government, then no more Joe Stacks. (See Stack’s interpretation of Section 1706 of the Tax Reform Act of 1986, which got me, too. In fact, I’m surprised there aren’t more software engineers flying airplanes into buildings.)
But now let’s return to Johnson’s DN interview, because here’s where my questions arise; where we apply an open source critique to the ObamaCare debacle. DN, and radio as a medium, perhaps aren’t the ideal mechanism to convey complex technical ideas. So let’s take this as the lowest common denominator expression of a process the public should fully own. I’ll key material where I want to ask questions — and no, I don’t feel the need to break out the color coding markers! — with numbers in square brackets, thus: .
AMY GOODMAN: … What do you think happened? Why has this website not worked?
CLAY JOHNSON: Well, government doesn’t have a lot of people to choose from when they’re looking for contractors to build this stuff. And I think part of the problem is that the same people that are building drones are building websites. When government is building a website like this, they have to use a system called procurement, which is about 1,800 pages’ worth of regulation that all but ensures that the people who are building this stuff are the people with the best lawyers, not the people with the best programmers. And so, you know, you have this sort of fundamental lack of talent amongst the contractor ecosystem that’s building this stuff, that it’s bound to be bad work—that, combined with the fact that in 1996 Congress lobotomized itself by getting rid of its technology think tank, called the Technology Assessment Office. So when they’re writing bills, they don’t understand the technology that they’re requiring in their laws. This is what you get when you have a Congress that is basically brainless on technology, and government who can only pick from a few old, stodgy contractors. You’re bound to have this result. And, in fact, the standings group came out earlier this week and pointed out that over—for all procurements over $10 million, 94 percent of them fail.
 “The same people that are building drones.” Yeah, Lockheed, ouch. (Though surely not a Canadian company like CGI?) And technically, aren’t drones innovative, and don’t they work pretty well? It’s not Lockheed’s fault Obama’s drone strikes keep blowing children to pink mist. That said, does open source scale? ObamaCare is a big systems integration project, and comprises Federal Departments like DHS, IRS, CMS, the Department of Indian Affairs, the Peace Corps, and Medicaid, at least one credit reporting agency (Experian), and 36 states. That’s just big. Do we have a precedent for a project of that magnitude being executed successfully as open source?***
 “People with the best lawyers.” As always, complexity and excess lead to niches for rentiers — ObamaCare itself being a prime example of this, except that the rentiers are government contractors and not the health insurance industry — and in procurement indeed, the “Beltway Bandits” leverage their political clout, knowledge of the process, contacts in government, and even existing contract vehicles to get work they should not get on technical merit alone, at prices that go far beyond the value of the productive work that they do. That said, it’s always possible to make things worse. How many pages of regulation should there be, if 1800 is too many? Does the UK’s Digital Government Service provide a precedent? And language like “work with agencies to make the right decisions” really gives me — as a taxpayer and a citizen — pause. (“If men were angels, no government would be necessary.” ) If I take that language literally, and translate it to the local level, it would mean getting rid of those pesky building codes and zoning regulations and letting the Town Office and the real estate (not software) developers “work together” “to make the right decisions.” That would lead instantly to open corruption and a destroyed ecosystem: decisions that benefited the developers only, and nobody else, and things are ugly enough as they are. Now, surely Johnson doesn’t mean for the language to be interpreted as I just did. But what does the language mean?
 “Lack of talent.” Diplomacy aside, I’m just not sure about “fundamental lack of talent.” When I was a contractor — albeit at a far lower level than Johnson — I met plenty of smart, committed people in government, and when I worked in a corporate cube I met plenty of, er, drones. Surely the programmers running the Mars Exploration Rovers are talented?
 “Lack of talent” “leads to bad work”? But what does “bad” mean? What if the incentives are structured to make “bad” work more profitable than good? Consider self-licking ice cream cones:
As far as I am aware, the phrase came into usage among knowing observers of the Vietnam War. A self-licking ice cream cone is a programme or policy that costs money and resources, generating a great deal of activity; produces indicators of its own success, preferably quantitative; but does not actually achieve its announced goals. Indeed, a proper self-licking cone undermines the very purposes for which it was created, while at the same time sucking in ever more resources from worthy and effective activities.
In the early 1960s, the US Air Force (USAF) advisory mission was trying to expand its role in South Vietnam. For this it required targets that were helpfully supplied by South Vietnamese officers. The targets were labeled “VC arms factory” or “command and control facility”, but were in fact peasant huts or the most prominent building in a village. The fighter bombers would go out and blow up these “structures”. Damage assessments duly reported a high rate of “success” – blown up huts.
To really grow this cone and set the self-licking in motion, an ever expanding list of targets had to be invented.
More targets meant more bombing, and more bombing meant a bigger role for the USAF. The result was more budget, more planes, and career advancement for all concerned.
The only problem was that the bombing was helping lose the war. Predictably, the peasants turned against Saigon, leading to the introduction of US ground forces in 1965. Like the “body count”, the number of “structures” destroyed had no relation to actually winning the war. It was an indicator of organisational “success” wholly divorced from reality, and from the values that the organization was supposed to be serving. The USAF was killing the very peasants it was there to save from communism.
Self-licking ice cream cones can arise in any modern organizational environment. People dedicated to the values of their vocations always have had to struggle against them–as did more than a few US officers in Vietnam.
Surely self-licking ice cream cones are an issue in software procurement? After all, from a pure profit perspective, a project that takes a $100 million to fail, and then another $100 million to fixand maybe another $100 million on top of that is far preferable to a $100 million project that just works! Now, it is true that I can think of ways that Johnson’s open source proposals would mitigate the self-licking ice cream code dynamic: the code is open, so there are more eyes on it; the firms are smaller and there are more of them, and so there’s more competition (at least at first); open source programmers have a strong sense of vocation. Nevertheless, do not the incentives for self-licking ice cream cones still exist? And do not incentives structure behavior? Here is where a successful example at scale would be really useful.
 “Congress lobotomized itself.” There are lots of things Congress has not lobotomized itself about, the campaign cash proffered by a behemoth like Lockheed, which is their “understanding” of technology, being one of them.
 “A few old, stodgy contractors.” Johnson’s claim that the primary characteristic of contractors is that they’re “stodgy” and “old” strikes me as not thought through. Surely it makes more sense to treat them as corrupt rentiers? Follow the money!
 “This result.” There’s a noteworthy absence here: Like most, Johnson tiptoes around the issue of White House involvement (some would say “political interference”). The White House political operation set the fundamental parameters for building the system: Adopt the Heritage approach and preserve the private insurance industry. That must have translated (depending on methodology) into a complex set of requirements for a complex system. The White House surely set the schedule over the three years since the ObamaCare legislation was passed; the White House set the baseline expectations for success or failure of the project at launch; the White House ran the project under conditions of extraordinary secrecy; the White House was responsible for which requirements got triaged (and which did not); the White House set a drop-dead launch date; the White House changed the web site’s plan selection form four months before launch. We have examples of great success for the open source movement in campaigning; but as we all know too well, campaigning is not governing. Do we have examples of open source successes, at scale, in a context of “political interference”?
 “94 percent of them fail.” IIRC, an extremely large percentage of private software projects fail. Are there statistics that show what percentage of open source projects succeed?**** On the scale of ObamaCare? In a context of “political interference”? Presumably there are not, or else Johnson would have supplied them.
Summing up, I really don’t want to rain on the open source parade. I really don’t. I use open source software all the time! And I’d rather live in Johnson’s world, where there are lots of small businesses writing open source software for the government; lumbering dinosaurs lose and agile mammals win!
However, I’m not sure that the technical issues that Johnson raises, though interesting, are the most interesting. The ObamaCare debacle took place — and any government program must and will — at the intersection of technology, policy, and politics, and we might ask ourselves which of those three forces is the driver. Clearly, policy drives technology; if, for example, it’s the policy of the United States government that only bi-planes will be built, then two-winged technology will at length issue forth from the procurement process. Equally clearly, politics drive policy; if it’s policy that only wooden bi-planes be built, and we find that bi-plane production is more or less evenly distributed among all the congressional districts where forest products dominate, then politics is where we’d look for explanations (“A Wooden America Is A Strong America!”). Bringing this home to the health care debate, which would be preferable: An efficient and Agile implementation of the complex ObamaCare exchanges, or a brain-dead and “stodgy” implementation of a dead-simple single payer system? Clearly, the latter — even if that’s a policy/political question — especially if your concern is excess fatalities among those not covered by ObamaCare even when it’s fully implemented. And if you’ve saved money by doing something that you should never have done in the first place, have you really saved money?
A political economy that was capable of procuring software as Johnson so justly would wish it to do would probably be capable of adopting single payer as well. I hate wicked problems…
NOTE * Definition and history of the term:
open source: n. [common; also adj. open-source] Term coined in March 1998 following the Mozilla release to describe software distributed in source under licenses guaranteeing anybody rights to freely use, modify, and redistribute, the code. The intent was to be able to sell the hackers’ ways of doing software to industry and the mainstream by avoiding the negative connotations (to suits) of the term “free software”. For discussion of the follow-on tactics and their consequences, see the Open Source Initiative site.
See also the definition at the Open Source Initiative.
NOTE ** If I were king, I’d give preference to small businesses that were also co-ops.
NOTE *** When I look at the projects listed at the UK’s Government Digital Service, there’s nothing remotely of the scale of ObamaCare.
NOTE **** I don’t know whether the definition of success for an open source project is the same as the definition for a project with proprietary source, whether corporate or government.
NOTE I may have more to say about this when I look at Krugman’s column tomorrow.