Yesterday, I was on a panel at The Server Side Java Symposium at the Venetian in Las Vegas. It’s been 7 years since I have been to the Venetian when we had the Documentum user group meeting, Momentum, at the Venetian. Having since been to the real Venice and been in a real gondola, the effect of seeing indoor canals and tromp l’oeil ceilings is even more surreal. (BTW - Don’t pay 80 euros to ride in a gondola in Venice. Spend one euro to take the Targetto. Don’t spend any money to ride a gondola in Las Vegas.)
This is the first time that I had met Joe Ottinger, the editor of The Server Side, in person. We have spoken on the phone before and he is a nice, funny, unassuming guy. Joe was the moderator of the panel that included Joaquin Ruiz from Spike Source, Brian Kim from Liferay, Bob McWhirter from JBoss.org, and Neelam Choksi from Interface 21 (Spring Framework). Geir Magnusson of Apache was also planned to be on there, but had other commitments. I know Joaquin and I were a bit surprised when we found out that the panel was only supposed to last 35 mins. (I got up at 5am and flown all the way to share 35 mins with 5 other people?!) Fortunately, Joe was able to start the session earlier and let the panel run on a bit longer. Most people stayed as well for the post-lunch session.
Most of the questions were about why open source and what license people should use. The reasons were consistent among us around the power of the model, the cost effectiveness of open source and the draw of the community. Joe had a question on how the community is managed and I think it was Dave who answered that it requires a certain level of control and discipline to ensure the quality of the code that is contributed. There were lots of questions around GPL, of which Joe is not a fan nor is he a fan of Richard Stallman. However, I think how people perceive GPL has changed in that you can combine non-GPL components with GPL and the FLOSS exception allows you to embed GPL in non-GPL systems. Everyone wants to know what will happen with GPLv3 and although we are tracking it, I can’t really say what will happen with the restrictions. (See Matt Asay's comments on GPLv3. Matt is our VP, Bus Dev and he knows a lot more about this than I do.) I commented on the concern that some OEMs have with LGPL in that it has not been tested in court. JBoss and RedHat have a very liberal and ISV friendly interpretation of it, but we need to have a court test to prove that it is okay to use in commercial software.
Richard Stallman
I appreciate the time that Joe gave us and the opportunity to speak about Alfresco. The Server Side has been a good source of the downloads that we have had so far. However, the whole area of Open Source Business Models is so much more than licenses. It really is a disruptive model not just for software, but for many other industry sectors. I organized a few thoughts on the flight over to Las Vegas.
Peter Fenton
When we first started to look at starting Alfresco, we interviewed and discussed the concept with some of the best thinkers in open source. One person in particular was very helpful due not just to his exposure to open source, but due to the thinking he had applied to looking at open source in the abstract as an investment thesis. Peter Fenton was an associate at Accel Partners and now a General Partner at Benchmark Capital. Peter had come up with a set of criteria of what worked and what did not work in open source. At the time we started, there had been 10 years of experimentation in investing in open source, although I think the last couple of years have been particularly instructional. Here are Peter’s criteria:
- There must be a large market with millions of users. I would characterize this as having a critical mass. If you are going to try and spread this as far and wide as possible and you are only going to get a couple of percent actually converting to paying customers, then that number must be very large. In this case, horizontal markets are good and vertical markets can be very limiting.
- There should be organic adoption of the project. As Peter points out most projects flame out in the first 2 years. The long, slow adoption of the organic community is a good thing. However, there are other ways to get a community. We have found a natural community in Enterprise Content Management practitioners who find Alfresco familiar and with which they are able to deploy much more cost-effectively than closed source options. Likewise, companies like IBM and Adobe have ready made communities as they exploit the open source model. Perhaps Microsoft will one day realize the potentially powerful open source community that they will be able to create with a quick email from Bill.
- There should be demand-side economies. It’s hard to know exactly what Peter means here and it might be Stanford MBA speak for scarcity of open source alternatives. He states that there should be a natural monopoly or duopoly. As an investment, you are looking for a company to be big and therefore it needs to be not just a leader in the segment, but dominant. Open source web content management on its own is an example of a market that is too diffuse. Open Source Enterprise Content Management is not.
- There needs to be sufficient product complexity or drift. This complexity provides a means to add value-added service over a product that is free. Products like application servers and database management systems are inherently complex. Many of the Apache projects, such as Tomcat, are too simple for someone to make money on them. It is possible that a product can work too well.
- The project should address a new frontier of adoption. Innovation happens at the edge. This should not just be old stuff for free. Peter uses the example of the LAMP stack as addressing new web sites and an area that commercial vendors overlooked. This is classic Innovator’s Dilemma. However, I would state this as there being a compelling differentiation other than just being open source. The 10 times factor of performance or new functionality allows the product to sell itself with the open source model being a swift closer. The community can be the source of the innovation that adds this wow factor.
I would add my own corollary to rule number one. Rather than just a large market, it needs to be a commodity market. With the commodity market you get the size need for critical mass and the market has been educated on the need for that product. This, by extension, means that not all markets are ripe for open source. Lall stated during the panel that not all markets need to be commodities since open source has been a major source of innovation. I agreed that some categories of open source were created by open source, but they are still commodities. An example is wikis, of which there are lots and it is now a commodity.
Looking at these rules, there are a number of industries that fall into these categories. Back in 2005, Peter was evaluating content management, wireless, system management, internet browsers, security, business intelligence, application infrastructure, middleware, databases, build and test environments, development tools. The only major category that was missing was games, which has 3 of the top 20 software companies. Most of these categories have had investment and new entrants funded by the major VCs, including Peter while at Benchmark. Perhaps you can consider other industries that are commoditized where you can tell one product from another except through very deep analysis.
Tim O'Reilly has said that when the price of a product like open source goes to zero, then the adjacent activities in the value chain gain value. This fits very neatly into Michael Porter's concept of value chains and competitive value. How you make money from open source can be divided into a few different models.Those models are dependent less on how you make money on the software and more on other activities that are adjacent to delivery and use of that software. The most common models are:
- Support - This is the model that Alfresco uses. This about providing technical support, maintenance and bug fixes. In this model, the free version usually moves faster than an enterprise version where more time is taken to certify the product against stack components. Bug fixes to this enterprise version are given higher priority, but are eventually folded into the main code line.
- Professional Services - This is providing professional services in using the software to create new solutions specifically for a customer. Custom engineering can also come under this category although the result can end up benefiting everyone as an enhancement to the product. JBoss and Interface 21 make some of their revenue from this model. At Alfresco, we give almost all the service revenue to our partners. Services is also one of the lowest margin businesses with the greatest risk in terms of building capacity.
- Insurance - This is a warranty or indemnity on the product. The company supporting the open source product guarantees to support the customer for damages or claims that may arise from the use of the product. This is rarely provided on its own and is usually provided as a value-added service in conjunction with either support or services.
- Enterprise Features - For those venturing into open source from a closed source background, this is the most comfortable model. By holding back certain features, particularly those that are related to scalability or system management, then users can try the software out, but if they are getting significant value from the software, then they will pay for this. We tried this model, but have decided that we could build a much larger community and hence more customers by providing the product 100% open source. Having enterprise features also encourages the community to replace your enterprise features, which can end up being counterproductive for everyone.
- Hosting - Hosting is becoming a more popular way to consume open source software. The people who create the product can get value by providing the software as a hosted service.
Embedding - Commercial vendors who can use open source software can often license the software to include it in an OEM relationship. This model is often accompanied by a dual license, a license that can be distributed as only open source or as an exception to those who pay for a commercial license. Once a commercial license is purchased, then the licensee is free to use the software however they feel appropriate. This is a model that was pioneered by JBoss, but originally developed by Ghostscript, an open source PostScript interpreter.
Each of these models has different implication in terms of how big the community it creates, the perception it gives to those trying the software and how much pressure there is on users to convert to a paid-for version of the software. Those who have given trust have often returned it with either contributing back to the project or paying for support when deploying the product into production. The exception to this are Google and Yahoo, who have created multi-billion dollar businesses using open source, but have given nowhere near as much back. This is one of the reasons that there is so much controversy over GPLv3, which attempts to set boundaries on fair use in a hosted environment.
The open source model is very powerful and rapidly growing. In our first full year of revenue, we have grown 50% faster than Documentum did in its first full year of revenue. You can use this model regardless of where you are located since it is inherently global. Despite slower uptake of open source in the UK, we have been able to distribute our software to countries where open source has greater acceptance and make a big impact in the US, the largest software market. It’s also a lot of fun, because people who are buying your product are doing it because they have already tried it and like it, rather than being convinced or coerced into using it.
There is still a lot experimentation in the above dimensions in the model, but I think we will be seeing a repeatable, cookie-cutter model. After all the fuss around GPLv3 dies down and after there is clear precedence on interpretation of LGPL, I think we will worry less about license and more about growing open source faster and seeing the disruption through.
The Mozilla Foundation has made Google Firefox's default search engine, earning a large but undisclosed sum to financially support it. Though 'embedding' comes close, I'm not sure if this well-known OSS revenue earning coup is specifically covered by any of your models above.
Posted by: Adam Pope | 2007.03.28 at 09:36 PM
I'm not so sure about that "commodity" idea. I think you're on to an idea, there, but its name is not "commodity." I've tried getting people to switch wiki engines, and I know better than even to TRY getting them to switch editors, and the reaction I get tells me these are not commodities (interchangeable).
I'm not sure what name the idea _does_ deserve, though. Maybe just "near-commodity" is good enough.
Posted by: Jack Repenning | 2007.03.27 at 04:29 AM