SourceFire circa 2002, purveyors of Snort via OSS and hardware appliance, now IPO'd? Not a startup.

Sleepycat software, purveyors of Berkeley DB under dual GPL and commercial license, acquired by Oracle? Not a startup.

Splunk, purveyors of Splunk, open source core and proprietary commercial extensions, with over 20MM in VC funding? Not a startup.

Hyperic, purveyors of Hyperic, which is OSS end-to-end, with a 6MM B from Accel? Not a startup.

Zimbra, purveyors of OSS Exchange clone Zimbra, acquired by Yahoo!? Not a startup.

Astaro, purveyors of one of the top 5 "UTM" (IPS+firewall) boxes, now profitable? Not a startup.

None of these companies are services businesses.

Your characterization of MySQL is laughable, too. The "cheap advertising version of MySQL" you mention accounts for 99% of the real-world usage of MySQL in the marketplace, and anybody on Hacker News wishes they had the market power MySQL did.

So, my response to you: name an equivalent number of companies any of us has heard of with an open-source product that is services-driven that isn't Red Hat or IBM.

You could add TrollTech, purveyors of Qt, to that list as well.

> The frequent cry of "but we'll make money selling support" ignores a critical fact: There aren't enough hours in the day to get rich by selling support contracts.

Speaking of ignoring critical facts:

1) You can hire more people to do support, which increases the total available amount of hours in the day.

2) Most support contracts are barely used and bought purely for comfort, so you can oversubscribe your support team to increase profit margins.

3) Every year software companies make billions on support contracts. MySQL made many millions on support.

Regarding 1) - yes... that's true, but it's a very linear model, which means it's of less interest to people like VC's. 2) is true, but lots of people realize that they don't need the support. The better you can bundle it, like Redhat does, the better off you are, because it means sticking more people with fees for something they don't need.

It's not an easy road.

Agreed. You should have a very proven path to generate revenue, outside of licensing, if you go the open source route. Companies who try to go it "halfway" by releasing a core product as OSS and then try to introduce closed-source addons have a tough time getting by on services while they try to prove that the addon is worth the extra cost: ArsDigita had to try this route in order to grow the business, and they burned through $20M before getting sold for pennies on the dollar to Red Hat.

In short, fantastic tools with which to build a business, but as a selling point of a product...once you take the genie out of the bottle, you often find yourself wishing you could put it back in later.

Phil Greenspun wrote at painful length about what killed ACS, and it certainly wasn't open source.

Now, provide another example of a company that screwed up by "halfway" doing OSS without "proving the addons were worth the cost". I've provided plenty of specific counterexamples.

Ha ha ha, how about Sun open-sourcing Java way too late to be of relevant use to anyone?

I'm more or less in agreement. The problem is scarcity - what you can make people pay for because it's scarce:

http://journal.dedasys.com/articles/2007/02/03/in-thrall-to-...

Looking at several successful "open source" oriented companies, we can see what kind of scarcity they introduced:

- Mysql, as you say, had their wonky licensing and a commercial product.

- Redhat decided to not license the 'branding' component of their distribution under a free license, so while you can copy the software, you have to remove a lot of bits and pieces (which CentOS has done), and it's not really Redhat anymore... If you want the support, you have to pay their licensing fees.

- Cygnus was relatively successful, a while ago, and were successful because they seemed, IMO, to stick to the letter of the law rather than the spirit: they'd give their clients full source code, but didn't make it public. They did a lot of 'billing' work, too, IIRC.

Still - I think there are a lot of advantages to using open source, and in many cases there are even business cases for developing it, but it is hard to base a business exclusively on open source.

If you make a development component, you can dual GPL/commercial license your product, like Sleepycat.

If you make an enterprise product, you can sell an appliance or a VM that runs it, and GPL the programs themselves, like SourceFire.

If you make a web product, you can GPL the code and sell the hosted version, like WordPress.

If you make an extensible product, you can GPL the core and sell the bells and whistles, like Splunk.

If you make an end-user product, you can GPL the core and sell the UI.

Good examples, which confirm what I said: you have to introduce an element of scarcity somewhere, that is not the software itself (with the possible exception of GPL'ed libraries, but that's sort of a special case).

When you make the decision to go open source obviously it's not from the product it self you are making money directly, only when sell licenses with support probably, but I am not excited with that model due to the fact that is not really clear how you can succeed reaching businesses to pay.

Thats my concern basically, while also the probabilities are mostly against you, and probably the product will disappear.

Of course, you seize to be considered a startup anymore, because a startup is about rapid growth.

So probably thats why te best way to go is to have a free version of the product and more advanced features for those who pay, and a seperate license to have access to the source code, like Symbian does?

But portions can be open source, as long as you keep what makes you as a business closed initially, you can retain control over your bloodline, while either open sourcing chunks or utilizing other open source projects.

Also: by your definition, Reddit and Octoparts? Not startups. Good call, Colin.

How does that follow from the above definitions?

I don't agree that if you build open source software, you aren't a startup -- but I do think that a big part of being a (VC) backed startup is the potential to achieve "hockey stick" growth, and that can be hard to do through support contracts, consulting, and certification alone.

"To be a startup, a company has to be a product business, not a service business."

Reddit and Octopart are product businesses, not service businesses.

By some tortured definition, sure. You win!

In what sense do are those sites "service businesses"?

The whole point of the service vs. product distinction that PG was making is that a service business is less scalable: its revenue is basically linear in the number of employees that it has, because a human being has to do more work for each new customer. That is not the case for product businesses or for Reddit or Octopart (or most other web startups) -- if Reddit's traffic increases 100x overnight, they don't need 100x the employees to capitalize. Craigslist is an example that makes the difference pretty clear. Whether you want to call them "product businesses" or not misses the point.

It's all a question of where you draw the line, I guess. There are consultancies, which scale 1.5x. There are service providers, which scale 2-3x. There are ASPs, which scale 5x. And there are product companies, which scale 10x.

Reddit is still different from an actual product company. The Reddit product scales independantly of the number of redditors or subreddits, because the Reddit product isn't hosted.

This is all a tangent, though. My fault.