Downloads, Adoption and Cumulative Advantage
Jan 13th, 2009 by sharps
According to Dimitris – our JBoss AS 5 downloads had a pretty big bump in December and digging a little into the SourceForge stats a few other things jumped out. First December 2008 was the second biggest month for downloads of JBoss AS (all versions) since 2003 and the downloads (all versions) have topped 100k/month for the last 4 months – that’s the best 4 months ever. January got off to a good start as well so hopefully the trend will continue. When you add all the downloads for the CR and Beta releases of AS 5.0 – it’s around 250k and that’s already more than 4.2.0, 4.2.1 and 4.2.2 (over much longer periods). So JBoss AS 5.0 is shaping up to be one of the fastest adopted versions JBoss has ever shipped.
So why should I care about downloads ? Well downloads are a pretty good proxy for adoption and adoption is key to realizing Open Source’s advantage. Every download is an opportunity for someone to try the technology and every time someone installs the technology there’s an opportunity for it to be used in a new and creative way and surface previously unseen deficiencies. And every time a deficiency is uncovered there’s an opportunity for the technology to improve. Multiply that by hundreds of thousands and add in rapid release cycles and it’s easy to imagine the cycle repeating as technology improvements drive greater adoption and greater adoption improves the technology. And that’s the cumulative advantage that OSS has over the traditional proprietary product lifecycle.
So go ahead – download AS 5 and see if you can break it.
How does JBoss compare to Tomcat, or Glassfish? I would love to know who is leading the charge to bring freedom to the enterprise…
Alicia, some previous posts might give a balanced view of the adoption of JBoss, Glassfish and Tomcat :
OpenLogic’s AppServer survey : http://blog.softwhere.org/archives/639
OpenSource Adoption – a survey of surveys : http://blog.softwhere.org/archives/558
Devox un-survey : http://blog.softwhere.org/archives/577
Specifically for bringing freedom to the Enterprise – JBoss did this in 2005 and as part of Red Hat is in a better position than ever to free users from out-dated, expensive, proprietary middleware. Obviously I’m biased as I work for Red Hat
- Rich
re: Comparison — Last time I did the comparison was in September; I posted the results at [1]. I need to update the comparison with the last few months of data but it has not changed substantially on the GlassFish side.
[1]http://blogs.sun.com/pelegri/entry/glassfish_download_stats_sept_2008
Eduardo, as part of Sun’s move to Open Source and greater transparency – when will Sun’s download stats. be made public ?
Also I wonder if Sun would be willing to plug JBoss downloads on the java.sun.com web site – as a long-time supporter of Java – we’d like to benefit from the association with Java as well as Glassfish
- Rich
re: when will Sun’s download stats. be made public ? — We have been doing that in an occasional manner. That is the URL I pointed to – I posted two other reports previously, so I’m collecting all of them below. Note they are all cumulative; so, although they are occasional, they show every month.
* http://blogs.sun.com/pelegri/entry/download_trends_for_glassfish
* http://blogs.sun.com/pelegri/entry/more_download_statistics_for_glassfish
* http://blogs.sun.com/pelegri/entry/glassfish_download_stats_sept_2008
I’ll do an update this week or next week. I don’t push them weekly because there is a fair amount of manual labor involved in validating the numbers. Yes, we should automate this. Yes, it is in our todo-list, but it is not prioritized at the top. Or do you mean something else?
re: JBoss downloads at java.sun.com — not my call; you know my bosses email addresses…
– eduard/o
Eduardo,
As you know, your stats include packages of Netbeans+Glassfish as well as Java SE+Glassfish (and possibly other form of packaging). Our stats above only list the standalone download.
Cheers,
sacha
Hi Sacha. My stats separate the “plain SDK” from the “tool bundles”. You can roughly compare SDKs to your standalone download. Not 100% but I believe that part is a pretty fair comparison. I explicitly separated the two portions to attempt a fair comparison
I am trying to make a fair comparison. I only used the SourceForge data for JBoss because that is all I had access to, but, as I previously mentioned to you, I am ready to add any other numbers that you think are applicable to the JBoss total. Fair comparisons are difficult with partial data; for example, we are counting only “completed” d/ls – I’ve not been able to find out whether SourceForge uses that same metric or “initiated” d/ls, and, if the latter what is their completion rate, which, in our experience, varies very substantially based on client locale, file size, and serving infrastructure.
– eduard/o
Eduardo, I was really interested in seeing the raw data – it just seems that it’s information that the community should have free access to. While the Sourceforge stats. aren’t perfect and a little unreliable from time to time – they’re open to everyone to scrutinize.
A big chunk of JBoss’ distribution is completely opaque to us – a lot of the technology like AS, Cache, Hibernate, Groups are embedded in other popular OSS projects and products.
I don’t think a direct comparison is really possible and wasn’t attempting to do that – the trends are more important to me.
- Rich
Good question on the boundary between transparency and competitive info.
In many cases we have chosen to be transparent even though we know the information can be used by our competitors; for example detailed roadmaps, detailed feature sets, geomap data, etc. In other cases we have decided to keep some information private (or more accurately, we have not decided to make it public).
Regarding raw d/l data, note that most of these downloads are d/ls of “Sun GlassFish Enterprise Server” distribution from Sun.Com, not the “GlassFish Server” distribution from Java.Net. These two distributions essentially the same binary, except for the name and a couple of small changes, but I think the Sun distro d/l numbers should be interpreted as Sun data, not community data, and at this point we are disclosing the aggregate but not all the raw data. I will bring up internally the topic of making public the raw Java.Net d/l numbers.
Regarding other distro’s of our AppServers, I agree they exist and are hard to count, but if you have any way to estimate their size, I’d be interested. GlassFish is in other distributions too but our belief is that most people just go get the latest version from our sites. That may be different in your case, don’t know, but part of why I started publicizing the GF numbers and I included the JBoss AS numbers from SourceForge is because I had found everybody was very surprised when we pointed out how many d/ls we had.
BTW, I need to update my charts for a customer visit on Friday, so I’ll post them at my blog and will add a pointer to it here.
– eduard/o
Eduardo, and I’ll be sure to add a long string of comments plugging JBoss to your future blog posts
Humm. Note that my initial post (#3) was very short, just 3 lines + URL. The rest is just me replying to you and Sacha’s post. If you think they were inappropriate, I’ll be extra careful next time. – eduard/o
[...] response to my recent post on JBoss AS 5 downloads by Matt Van Bergen of [...]
[...] that people expect takes a long time to attain – multiple releases over many years. I think mass-adoption speeds up the process and shortens the refinement cycles but even so – it takes a long time and it takes real customer [...]