Ohloh Report on GlassFish and Open Source Adoption


We just posted a new report on AppServer adoption that we commissioned from Ohloh (recently acquired by SourceForge). The report uses the extensive code analysis done by Ohloh (38,000 projects across 3,500 source forges) to compute several metrics on how Open Source projects are targeting different AppServers.

Ohloh computed trends through their application-specific deployment descriptors and found that ~50% of the projects currently target GlassFish and that the number jumps to 73% when only counting new projects.

Check out full details at the report and also see John's writeup.


Looking at the scanned descriptors, the ohloh report should really take into account 'jboss-web.xml' & 'jboss-client.xml'. But then, jboss didn't pay for it, right?

Posted by John McDonand on June 03, 2009 at 07:19 PM PDT #

Thanks, John, I'll pass that along to the ohloh folks. If anybody else sees other areas where this can be improved, please pass it along.

Unless these descriptors were introduced recently, it should not impact the trends, which I think is more important than the actual numbers.

Posted by Eduardo Pelegri-Llopart on June 04, 2009 at 12:51 AM PDT #

They are both old descriptors. I wouldn't expect jboss-client.xml to have much impact, but jboss-web.xml probably yes because it indicates a web application. Nevertheless, the trend looks clear.

Posted by John McDonald on June 04, 2009 at 02:06 AM PDT #

I don't think looking at descriptors gives an accurate picture at all.

Hello, annotations?...

I realize that the likelihood that the number of JBoss developers that are using annotations more being significantly higher than the number of Glassfish developers is low so the trend is probably still similar but we don't know.

This is simply a big hole in the analysis that prevents us from really drawing too many firm conclusions.

Not including JBoss web descriptors also probably fits in this category. Of course with that the likelihood that JBoss developers were abandoning EJB entirely around the time JBoss started supporting EJB 3.0 is a bit of a stretch but that is the argument the Spring camp would make.

I would be willing to overlook the omission of annotations in the analysis as long as JBoss web descriptors are included.

Now don't get me wrong, I am really excited about GlassFish and am going to take a serious look now that JEE 6 is finalizing. The fact that JBoss couldn't even beat WebSphere in implementing JEE 5 is just a joke!!!

My company has also done a lot of JBoss administration training with our clients and how difficult this is with JBoss compared to Glassfish or even Apache Geronimo/WAS CE is another joke.

The fact that clients haven't paid for JBoss means they are more likely to ditch it if something better comes along. I haven't seen anything with JBoss 5.x that would lead me to suggest our clients stick with JBoss so I'm starting to suggest they switch. Glassfish will likely be my choice for JEE 6 implementations.

Great stuff!
Stuart Smith
Administration Lead
Web Age Solutions

Posted by Stuart Smith on June 04, 2009 at 03:03 AM PDT #

Hi Stuart - I'll see if we can ask Ohloh to re-run the analysis w/ the additional web descriptors but I'll wait a bit to see if other people point out missing descriptors.

I was not involved in commissioning this report; I'd love to do some data-mining on the Ohloh data to get a better picture of the situation and see if there is other useful adoption data in there. Based on what I see, the data seems self-consistent and in line with other adoption indicators.

Posted by Eduardo Pelegri-Llopart on June 04, 2009 at 03:43 AM PDT #

Thank you for the information.

Posted by Download on June 04, 2009 at 10:42 AM PDT #

I agree that it provides support for the statement "adoption of Glassfish is increasing dramatically". As I said, I'm more excited for the future of Glassfish than anything coming out of JBoss anymore.

Typically adoption might be measured with number of downloads. Analyzing the source code of open source projects provides another angle.

Open source projects decide which servers to target based on a higher level of technical evaluation than you might see for end users. End users are often constrained by other considerations than technical merit. Since open source projects start with a cleaner slate this metric might actually say more about the technical merit of a server than other measures.

Thanks for looking to follow up. It would sure be nice to have an independent analysis to share with clients that can't be dismissed.


Posted by Stuart Smith on June 04, 2009 at 11:44 AM PDT #

jboss-service.xml is another useful one to scan for as well.

Posted by Joe Khoobyar on June 05, 2009 at 09:38 AM PDT #

Post a Comment:
Comments are closed for this entry.