Dries Buytaert

Over at his weblog Earl Miles wrote:

The user base for Drupal has been growing exponentially, and even though the number of people who stick around and are willing to spend their time giving support are growing at the same pace, it's an exponential comparison, not a linear one. Which means the gap is growing. Compare the graphs of 10^n and 2^n. Where n = 1, 10 users per 2 support people, that is 20%. But where n = 8, that is 10,000,000 users and 256 support people, or .002% (give or take a decimal point).

I disagree with this assessment and the mathematical projections. I think that new developments, and not a growing user base, are the main cause for the quality of community support to degrade. Allow me to explain ...

To me, it looks like community support follows a Zipf distribution. If we apply Zipf's law to the Drupal documentation, it says that in a distribution of support requests, there is a huge number of requests for a small number of answers and a small number of requests for a large number of less common answers.

Applied to the Drupal handbook, this means that a relative small number of pages in the Drupal handbook, satisfy a large fraction of our users. People who don't find a quick answer in the Drupal handbook, might post a support request in the forums. Here too, Zipf's law applies. A huge number of support requests can be answered by a large number of people, and only a fraction of the forum topics can be answered by one or two people in the Drupal community.

As our user base grows, more and more questions get answered and more and more questions go unanswered. Fiddle with the parameters of the mathematical equation and you'll observe that as the popularity of support requests grows, their relative popularity remains the same. In other words, if Zipf's law applies to community support, the quality of support won't degrade as the community grows.

However, that is not what people perceive. People will focus on the increasing number of support requests that go unanswered and incorrectly conclude that the quality of support is going downhill. The truth is that, while we see more support requests, we are also successfully answering an increasing number of support requests. Thus, the problem is one of perception. As the user base grows, the long tail of Zipf's distribution becomes more visible ...

Clearly, these observations only hold for a static project. Drupal is a dynamic project as the community continues to add bells and whistles to the software. By doing so we introduce new support requests to the system. These new requests alter the relative popularity of the existing requests in the Zipf distribution, and therefore, can impact the quality of support. So what does this mean? It means that as long all developers do a good job documenting all the new aspects of Drupal, the quality of Drupal's support can only improve. It also means that if developers don't document new aspects of Drupal, or document them poorly, the quality of Drupal's support will degrade. At all cost, new developments and documentation updates need to progress at the same pace.

Anyway, my point is that (i) it is not the growing user base but the growing developer base that is to be held responsible and that (ii) the problem is one of perception because the relative quality of support can easily remain constant. Of course, perception matters a lot and it is not something we can ignore. After all, the (perceived) quality of support will be a key differentiator in the future. To solve the perception issue, we need to make Drupal easier to use so we can turn new users into contributors more easily. The lower the barrier, the faster you're an expert that can help others.

Like it or not, but with 2007 around the corner, it is almost reflections and predictions time again. With that in mind, you might find the graphs below to be a source of inspiration.

The past growth in absolute numbers. The "January 2007"-kink is due to the fact that data for December 2006 is not yet available at this point.

The past growth in relative numbers. In 2005, the number of nodes and comments grew by more than 300%. In 2006, the number of nodes and comments grew by more than 230%.

A growth projection using a polynomial fit (multiple regression). Simply put, the correlation coefficient R-square, is a measure for the quality of the fit. R-square can assume values between 0 and 1, where a value of 1 indicates a perfect fit. Bear in mind that a statistical fit may be anything but an accurate prediction of future growth.

When Steve Jobs had to fix Apple in 1997, he said the following: "The products suck! There is no sex in them anymore!". Well, we added so much sex to Drupal 5.0 that I'm horny all day. Tasteless jokes aside, working with Drupal 5.0 has a very strange effect on me. Like it helped me re-discover my love for Drupal -- except that this love was never lost. A strange feeling that is hard to explain ...

You can't sell Drupal, or any modification you made to Drupal. You can charge money for having to make these changes but you can't make these changes available under a commercial license. Why not? Because Drupal's license, the General Public License 2 (GPL 2), mandates that all modifications also be distributed under the GPL.

But when you are providing a service through the web using GPL'ed software like Drupal, you are not actually distributing the software. You are providing access to the software. Thus, a way to make money with Drupal is to sell access to a web service built on top of Drupal. This is commonly referred to as the web services loophole.

Some people say this loophole is inconsistent with the values of the Free Software movement. Others think of this loophole as an interesting feature. I'm in the latter camp. In fact, I predict that 2007 will bring a small tsunami of Drupal distributions built around a hosted service model.

Fact is, when the GPL was created 15 years ago, it did not really predict a world of web services. Version 3 of the GPL, expected to be released in 2007, will tackle this issue. It will allow developers to add an optional clause to the license that requires hosted service providers to share the source code and their modifications. This optional clause can be found in section 7.b.4 of the most recent draft of the GPL 3 (subject to change):

Additional requirements are terms that further constrain use, modification or propagation of covered works. This License affects only the procedure for enforcing additional requirements, and does not assert that they can be successfully enforced by the copyright holder. Only these kinds of additional requirements are allowed by this License: ... snip ... 4) terms that require, if a modified version of the material they cover is a work intended to interact with users through a computer network, that those users be able to obtain copies of the Corresponding Source of the work through the same network session.

Personally, I don't see us adding such a clause to Drupal. It doesn't bode well with the way people use Drupal, or any content management system for that matter. Here is just one example: most people theme their sites by downloading and modifying one of the available GPL themes. If we were to add the additional clause, this would no longer be desirable, because you'd be forced to share all your modifications, including your theme's. It is problematic when you want to create a unique site or brand.

So long live the web services loophole!

One of the most popular and deadliest vodcasts on the internet, Ask a Ninja, is using Drupal. This is what the Wallstreet Journal has to say about Ask a Ninja:

Each week, about half a million people watch a two- or three-minute video starring a man in a ninja costume that includes a Lycra ski mask bought for $6. He typically delivers a sarcastic comic monologue in response to a ninja-themed question a viewer has e-mailed in. ("Do ninjas catch colds?" was a recent topic.)

They have been using Drupal for more than 6 months now, and I have been watching their episodes for at least as long. Let's hope the Ninja continues to be happy with Drupal. Or doesn't the Ninja kill Drupal developers? Maybe that is a good question to ask?

Be careful though. When a question is sent to the Ninja, the following auto-reply is sent: The Ninja will read your question carefully and then decide whether or not it is interesting and funny. If it is, he'll answer it and look forward to killing you soon. If it isn't, he'll still kill you, but he won't enjoy it.

According to the Telegraph, Ecademy, a business networking site for professionals, and one of the first big Drupal sites, is for sale. Of course, for the right money, any business is for sale. Still interesting because Ecademy forked Drupal 3.0 in 2001 and has long been an important Drupal showcase.