World's biggest open-source company? Google

(Credit: Open Source Initiative)

Red Hat is generally credited as the industry's leading open-source company, but it's a distinction that is as meaningless as it is incorrect. While Red Hat's revenue directly derives from the open-source software it develops and distributes, other companies like Sun, IBM, and Google actually write and contribute far more open-source code. It may be time to stop talking about open-source companies and get back to the importance of open-source code.

Open source is increasingly the foundation upon which software and Web companies depend. MySpace made waves on Tuesday by open sourcing Qizmt, a distributed computation framework (running on Windows Server, intriguingly) that currently powers MySpace's "People you may know" feature. But MySpace, as VentureBeat notes, was simply playing catch-up to Facebook's recent open sourcing of Tornado.

Neither move is an attempt to score brownie points with the "in" crowd. Both moves are motivated by self-interest, self-interest that increasingly requires inviting developer communities to embrace and extend one's Web services/software through open source.

It's also a way to improve software quality. By embracing open-source projects as a foundation for a company's software, and then extending it through its own open-source projects, the collective quality of open source is strong and growing, as Accenture's Kit Plummer notes.

It's this enlightened self-interest and the quality is engenders that has turned open source into essential infrastructure for virtually all commercial software, and which means Red Hat and other pure-play open-source companies are no longer the center of the open-source universe.

The Linux kernel is comprised of 11.5 million lines of code, of which Red Hat is responsible for roughly 12 percent (measured in terms of lines of code changed). Even if we add in JBoss Application Server (another 2 million lines of code or so) and other Red Hat projects, we're still left with far less open-source code from Red Hat than from others.

Take Sun, for example. Sun is the primary developer behind Java (more than 6.5 million lines of code), Solaris (over 2 million lines of code), OpenOffice (approximately 10 million lines of code), and other open-source projects.

Or IBM, with 12.5 million lines of code contributed to Eclipse alone, not to mention Linux (6.3 percent of total contributions), Geronimo, and a wide variety of other open-source projects.

Google, however, is the most interesting company of all, as it's not a software company, per se. I asked Chris DiBona, Google's open source and public sector program manager, about Google's open-source contributions. His response:

Conservatively, we've released about 14 million lines of code. Android tops 10 million lines of code, and then you have Chrome (2 million lines of code), GWT (300,000 lines of code), and about a project released every week over the last five years. Then you have a couple hundred Googlers patching on a weekly or monthly basis.

While DiBona was quick to suggest that Google doesn't claim the crown for Open Source Top Contributor ("We'd say we're 'among' the largest [contributors]"), it almost certainly is the world's largest open-source code contributor, especially when one considers its other open-source activities, including hosting perhaps the world's largest repository for open-source projects, with more than 250,000 hosted projects, at least 40,000 of which are actively contributed, not to mention its Summer of Code. After all, lines of code, while useful, is not necessarily the best measure of the value of open-source contributions.

In fact, Patrick Finch of the Mozilla Foundation speculates that Google's best open-source contribution may have nothing to do with writing new code at all:

Google's biggest contribution to open source is arguably not code, but proving that you can scale Linux on whitebox hardware.

It's a great point, and one that underscores the fact that the "open-source company" distinction is somewhat useless. Google doesn't call itself an open-source company, and rightly so. Open source is simply part of its strategy for distributing software that will help it sell more advertising.

Sun attempted to turn itself into an open-source company, but once Oracle completes its acquisition of Sun, Oracle certainly won't take on that label. Not because it's a bad label, but because it's simply not a useful one anymore.

We are all open-source companies now. Which also means that none of us are. Open source is simply a way that we enable some aspect of our businesses, whether we're Red Hat or Microsoft or Google or Facebook.

And given that Web companies like Google don't need to directly monetize open source, we may actually see far more open-source code emerge from these Web companies than we ever have or ever will from traditional "open-source software companies" like Red Hat, MySQL, or Pentaho.

[jrepenning]

I agree that any term that describes everyone is pretty much meaningless, but I don't agree that all meaning has left the "open source company" world. Things have indeed changed, but there are still important things to be said. I'd say we need some new terms. Here are some efforts:

- A term for "company that uses open source software" hardly needs inventing any more. It's only remaining use would be the question "isn't everyone?" So, we'll skip that one.

- Some companies foster and contribute to open source projects in order to create their own primary product. Maybe call that "commercial open source," which has such sub-categories as "dual licensing," "fee for support," and the like.

- Some companies foster and support open source for the traffic (ad impressions) it brings to their hosting sites. Maybe call that "media open source."

- Some companies foster and support open source extensions to their primary product. Maybe "ecosystem open source"?

- And some companies foster and support open source products because they're necessary components in order to create the company's primary product. "Infrastructure open source"?

Many companies do a mix of the above, of course, but the terms are therefore all the more useful, since without them there's no way to say that.

[Matt Asay]

It's a good suggestion, and one that calls out the need for us to be much more descriptive in how we use open source (since everyone will be using it). Thanks for the comment.

[rob1400]

Google is both an open and a closed source companies. Its crown jewel / secret recipe is in its server / service in the back-end that is totally closed. Can you imagine Google opening the source codes of search engine? or gmail server?

Google opens only its client software to encourage the use of its closed services. That is its business model.

[rob1400]

Google doesn't mind opening the source of its Chrome browser, Android phones, or the upcoming Chrome OS. But when it comes to its backend, never think about it :)

[jessiethe3rd]

Google is open until it gets to the revenue generating business... that open story soon turns into a closed story. One could argue that companies like Oracle, Microsoft, and IBM have embrassed Open Source in much the same way - gateway to more revenues.

Companies are all in there to make money. Fact of the matter is Open Source does not equate to free or in many cases cheap or cheaper then commercial sofware.

[Random_Walk]

Not so sure about simply counting lines of code as a metric, Matt... Otherwise:

"void() { "
(no, that's the whole line)

...bears just as much weight as...

"multi sub infix:<?>(Vector $a, Vector $b) {"
(a snippet of a PD dot-product operator I scrounged up online just to illustrate).

Short version: there's no comparison of what it took to 'build' a given line.

[Random_Walk]

heh - I love how a superscript dot becomes a question mark after the website gets done mangling it :)

[kit_plummer]

Ha. This is the fundamental problem with all things SLOC, including estimation and valuation. And, of course the language obviously matters.

Trying to decipher the value of Bigs contributions is really a huge subjective undertaking. I believe the perception that an organization is "good" in the sense of community citizenship is what is really important. Right? To raise a ghost real quick...the is part of the "free loading" story. It is a give/take thing...so we really should be looking for a ratio that quantifies contributions - that come up with a formula that evaluates a notion of quality based on who else consumes the contributions.

Ok, I'll shut up now.

[eadeguzman]

Hmm... So what do you suggest? Removing all whitespaces and then count characters or "words"?

Nobody is solving for accuracy there. Lines of code is a sufficient measure. Things begin to be very tricky if you consider words and characters... as

String aVeryLongVariableNameHahahaha = "A lot of nonsense."

would count higher than

String av = "A lot of nonsense.";

Plus you have to deal with un-optimized code... meaning a lot of cut-and-paste of the same routine (instead of having a function)...

How can you measure those? Besides, I assume that dev projects would have some standards on how code is written and I'm assuming that these projects would undergo code review, etc.

[dinkeldorf]

True but over the size of code bases, it averages out.

[Matt Asay]

I agree that it's a poor measurement, but I couldn't think of a better place to start. I'd be very happy to get alternative measurements...if someone else wanted to do them for me. :-)

[pentest]

A better metric Matt is what is the value of the OSS that company X releases.

LOC doesn't imply anything, not performance(5 lines can be slower than 50 and both do the same thing), not features, nothing.

It is a metric used only by the lazy and ignorant, oh wait never mind.

[Anonymous44]

Matt, are you double-counting non-google-authored stuff in chrome & android? You do know that these are based on webkit & linux respectively, right?

[pentest]

Lines of code is a meaningless metric in any context.