Healthcare.gov Lines of Code Comparison | Hacker News Reader

Healthcare.gov Lines of Code Comparison

alexmarchant.com

62 pointsalex_marchant43 comments

Yeah newsreaders are really bad at single sourcing a quote without qualifying the source. And then repeating it. Endlessly.

That is basically how we got into the Iraq War, Cheney fed some nonsense to the NY Times which printed it, then Cheney went on the news the next week and quoted the NY Times and everyone started repeating the NY Times, when the original source of the nonsense was him.

By the way, is an opening/closing bracket on its own line a "line of code"? What if there are 200 million of them? Is that like double spacing your term papers in large fonts and wide margins to cheat the page requirement?

My general rule of thumb with unnamed sources is to completely disregard what they have to say. With named sources, my general rule of thumb is to assume they have ulterior motive for speaking to the press, usually apparent by looking at their job title. My general rule of thumb with all news is to wait for some distant period of time for the real facts to shake out of the mess of people talking.

"According to one specialist, the Web site contains about 500 million lines of software code."

And you guys believe that so call "specialist" without any proof?...What happened to HN? :/

It fits the current meme here that anything related to the US government technology wise must automatically be evil and/or incompetent.

Yup this is obviously bullshit.

If each LOC generated 16 bytes of native code, the executable would be 7.5GB.

It was good enough for the New York Times though?

It's disheartening to think they actually took that statement and printed it without verification. This simple chart illustrates how foolish that was.

They will print anything for views

Journalists are equal only to lawyers in their dislike of math.

All the threads about this 500m line claim has been filled to the brim with people highly skeptical about the claim. No, "HN" does not believe the 500m LOC claim

"According to one specialist, the Web site contains about 500 million lines of software code."

I call bullshit. Facebook is 62 million KLOC on the main site, excluding the backend. And I'll argue that Facebook is one of the most complicated sites out there, feature-wise.

I wonder if this is actually combining all 50 states' health exchanges (and their third-party insurance API connectors), which MIGHT make sense, given an average of 10 million lines per state site.

62 million KLOC = 62,000 million lines of code. I'm calling bullshit on that.

I presume they meant 62 MLOC.

> Facebook is 62 million KLOC on the main site, excluding the backend. And I'll argue that Facebook is one of the most complicated sites out there, feature-wise.

Assuming you mean 62 MLOC, I have no problem believing that the backend implementing federally-operated healthcare exchanges for the 27 states that didn't set up their own state-level exchanges is on the order of 10 times as much code as Facebook has when you exclude Facebook's backend.

> I wonder if this is actually combining all 50 states' health exchanges (and their third-party insurance API connectors), which MIGHT make sense, given an average of 10 million lines per state site.

The federally-operated exchanges at issue are "only" for 27 states; the remaining states operate their own exchanges. But, yes, obviously any count of the LOC of the federal backend is going to include the exchanges -- that's what the backend is.

It's probably combining the lines of generated HTML of every possible page permutation of the site.

I know you're joking - but seriously, I think this is it. I don't think it's "page permutation", but probably more like the number of page template lines of code. i.e. all the HTML templates + the embedded jsp/php/asp/whateverp embedded. Lots of people (technical management or marketers) likely think of HTML as "code" and thus it's included in the count.

I also think that the 500mm lines of code was likely floated to sound impressive - turns out, of course, this shows how messed up this project really is.

62e9 LoC?

I know the esteemed New York Times made the half-billion-lines of code assertion, but if you read the actual article, this claim is buried at the bottom. And this is the context:

> According to one specialist, the Web site contains about 500 million lines of software code. By comparison, a large bank’s computer system is typically about one-fifth that size.

http://www.nytimes.com/2013/10/21/us/insurance-site-seen-nee...

So "one specialist" figured it out? So who is this unnamed software-whisperer that the Times reporter picked out? It could be anyone from Linus Torvalds to an SEO expert, or maybe some doofus consultant who's an expert at getting gov't contracts and delivering terrible systems (on second thought, that person probably would know best).

If you consider the other articles written about this, which claim that contractors only started work anywhere from 7 to 12 months from now...how is it even physically possible to write 500 million lines of code?

OK, yes, you can do it with copying-pasting, which clearly was part of the problem (lorem ipsum makes a few appearances in the production JavaScript)...but who knows what that number actually includes? Just actual source code? Comments? If the system includes, technically, some distro of Linux, is that counted?

Speaking of the Linux distro, here's an article from 2008, referring to a study that found that Fedora 9 had roughly 200 million lines of source code, and was equivalent to about $10.8 billion worth of dev work:

http://www.linuxfoundation.org/sites/main/files/publications...

So hey, if we're treating lines of code as fungible quantities, it sounds like the Feds got quite the deal! (except for the part of it not quite being reliable)

As someone else pointed out there are about 380M people living in the US so perhaps the code is :

   if (name == "alice adams") {
   } else if (name == "betty adams") {
   ...
   } else if (name == "mark zuckerberg") {
   } else {
       open_dialog("Are you sure you are a US Citizen?");
   }

Sillyness I know. I actually agree with many people that someone likely pulled the "500M" number out of a hat during a stressful moment needing a "really big number" not thinking about what it meant if that was in fact the number.

It is also possible they are including the insurance provider database rows as "code".

Perhaps 317 million lines are a hardcoded list of every man, woman, and child in the U.S. That would leave only 183 million.

That 323 million lines of code for Debian 5.0 also counts all the packages included with debian. The debian codebase by itself is only 68 million lines of code: http://www.ohloh.net/p/debian

Good catch, I'll update this soon.

Looking through the source code posted elsewhere, it appears to make heavy use of AndroMDA internally, which appears to generate Java classes from UML diagrams. I think that's important to keep in mind when judging the 500m figure, as 500m computer generated code != 500m human code.

Have you found marketplace backend code anywhere? The frontend code is easy to find and there's at least one repo with it (see my previous code) but I'd be very interested to see backend code.

If you look at the code, e.g.

https://gist.github.com/adamnemecek/7112729/raw/deb521f699d3...

500 million LOC seems less unrealistic.

A simple `find . -name '*.js' | xargs wc -l` on the marketplace repository at https://github.com/STRML/Healthcare.gov-Marketplace reads about 72K LOC. About half of that is vendor scripts, such as jQuery.

The 500M figure is quite obviously nonsense. Even if one adds the source of every tool involved, including the browsers used to display the site, it would be difficult to get to 500M. For example, Chrome is ~5M LOC, depending on comments/version.

500M bytes, maybe? Including all the artwork, and rows of database gunk and XML propping it up?

I can easily believe some overflunky seeing 500M something and not really get what was being measured. I'd believe that more readily than 500MSLOC of actual code.

That is the FRONT end which is open-source. Made by different people.

The BACK end is closed source, made by CGI and others.

Of course, but I would be shocked if the backend was more than even 3-5 million lines.

The back end is the entire API.

But I tend to agree the size claimed has to be way off.

They might be including database rows as "code" who knows.

That code looks really familiar. Strangely enough, it was for another healthcare application and written by consultants.

This post is garbage and perpetuating this false statistic is useless noise.

Someone must be practicing enterprise coding. See also: https://github.com/EnterpriseQualityCoding/FizzBuzzEnterpris...

The entire site is generated via Jekyll with every possible permeation. I just made that up. But it does use Jekyll.

https://github.com/greggersh/healthcare.gov#install-jekyll

The front end does, that was built by Development Seed. The problems have seemed to come from other contractors, such as CGI Federal, though I'm loathe to trust anything I've read in the news about this clusterfuck so far. The reporting has been so bad it rivals Healthcare.gov itself in its awfulness.

The craziest part has been the hysteria about the healthcare.gov repository going down. It never contained any files that were useful to anyone - it was just the blog / landing page. It's like Microsoft promising to open-source their operating system and all you get is the software that generates the help files.

Why? Literally, how could this be? I think they must be counting every dependency twice.

Jesus, it's disappointing that HN is focused on this silly statistic instead of actually talking about the problem stalling millions of Americans right now from enrolling. Considering HN is kind of, like, the epitome of technical talent on the Internet. They would know best how to solve HealthCare.gov's problems.

https://news.ycombinator.com/item?id=6596981

seeds orange apple apple Apple applesauce [rooty tooty fresh and fruity breakfast]

import *; // so much work, almost 500 millions lines of code

Ah, it all makes sense now. For about a dollar a line, I wouldn't give a shit either.

...

HAHAHAHA!!!

Reply on news.ycombinator.com