Busting the Myth of GPL Dominance: Apache Rules!
So we all know that most open source projects are licensed under the GPL – Ken Krugle's graph shows that clearly over 75% of projects are GPL or LGPL – but it turns out that most open source software used by enterprises are not licensed under the GPL … companies are using primarily Apache licensed software!
OpenLogic has over 300 projects in our open source software library – added by customer request. (Our customers being large enterprises.) Taking a look at those projects I discovered that:
- 29% of them are licensed under the GPL or LGPL
- 35% of them are licensed under the Apache license
But wait! It gets even more interesting if you look at just the top 20 projects. I sorted our library by number of customers actually using the open source software package, took the top 20, grouped by license and found:
- 75% Apache
- 20% GPL or LGPL
- 20% CPL, Eclipse, Perl, BSD
(Note that this ads up to more than 100% because several projects had several licenses. That's why I didn't put these numbers in a pie chart – it would imply that each project just had one license.)
So now the interesting question is: is this coincidence or cause and effect?
- Do companies prefer the Apache license and so they tend to gravitate to projects licensed under the Apache license?
- Do companies have anti-GPL clauses in their open source policy that prevents them from using more GPL licensed software?
- Does the Apache Foundation (which is all Apache licensed) create better than average software and so that software tends to be more widely used? (Lots of projects on SourceForge are trial projects run by one person that never really went anywhere.)
- Is there actually more Apache licensed code out there than people have been counting because they usually run their statistics on SourceForge data?
- All of the above?
I'm voting on a little of all of the above. What do you think?
I would like to suggest another possibility. I don’t believe deployment decisions (for any type of software) are based on licensing terms and conditions very often. I think you are seeing the results of decisions made primarily because the enterprises needed products that just happen to be licensed under Apache. I have 30 years of research that says “Functionality rules!”
Can your statistics break out the Apache licenses by “Apache product” type? My guess is that a high percentage are classic Apache web server software, but that’s just a guess.
Also–if anyone knows–do you get an Apache license when you get Websphere and Oracle AS and other such products with Apache web server built in? That would skew the numbers also.
Dennis
Friday’s Recommended Reading…
What happens when you’re too busy during the week to blog? You get a backlog of cool reading material…….
Back in 2002, David Wheeler pointed out that about 50% of the software in the Red Hat distribution was under GPL.
Since the OpenLogic list is by customer request, what if the customers are mostly requesting stuff that they don’t already have preinstalled with their distributions?
Don, you are correct that we typically don’t support things that are part of the distribution already (since customers have another support path for those) but I don’t think that changes the Apache/GPL mixture.
Just wait a couple of years for corporate superstitious fears to wear off. The GPL will then be seen for what it is – the fairest license based on a win-win philosophy. The GPL is a pro-user license, but it’s ironic if it’s more popular today among developers than end-users. One can only conclude that when understanding deepens, it will attain its expected level of popularity among users.
Samba is missing, Linux Kernel is missing, what kind of customers do you have?
-_-
Well, I think there are some reasons for this – though not necessarily the ones give by you.
The reason one given by you (preference for ASL over GPL) ist not convincing. If so, there should also be a preference for the BSD-license over GPL because it is as permissive as the ASL.
I think reason three (better software – I would prefer the wording very good, though not generally better software) is pretty much valid. Apache has a very rigorous process before projects can become Apache projects. They must already have a viable community and developer base. There is no such limitation on Sourceforge/savannah.org/berlios.de or most other project aggregating sites.
I think another issue is important. You just examined large enterprise customers. These tend to prefer Java over other languages for their projects. And Apache has some very important projects for Java developers. Struts (okay a bit dated, but still heavily in use in large enterprises), Tomcat (by far the most common Servlet container), MyFaces, Ant, Maven, Apache commons and many more. And every Java project I have seen in the last five years uses some (as in plural) Apache libraries.
And of course the most used web server is Apache’s httpd. So I am not that surprised about your findings.
Stormy, David searched 612 packages from Red Hat, and 55% of those were GPL. If you have 300 at 20% GPL, someone who installs all of everything will have about 44% GPL by package count.
rittmey, you’re right about the Java effect. Apache has a lot of popular Java projects, and until Java went GPL and Red Hat bought JBoss, Red Hat wasn’t including them as part of its distribution. That created a niche for the “stack” companies to do for Java on the Java VM what Red Hat did for C on x86.
It’s a big mistake to lump GPL and LGPL together as if they were essentially the same. As far as developers are concerned, LGPL is much closer to Apache than GPL.
[...] Busting the Myth of GPL Dominance: Apache Rules! | OpenLogic Blogs [...]