Modeling: Home Star Loader

Last week I reported that there are now 70 committers / contributors in Modeling.

Today, Denis announced on the Phoenix mailing list that EMF, EMFT, MDT and GMF are in the top 30 requested home pages for Feb 2009.

The top three popular project pages were, in order: PDT, EMFT, and WTP.

However, if you combine the numbers for the four Modeling subprojects, you get 162K hits for the month, placing Modeling in the #4 slot behind only the Eclipse.org home page, Downloads page, and Categories page, and making us the #1 (aggregate) project page.

Download Stats & Missing Data

According to Google Analytics' new "snot graph" tool, the top three downloads in Modeling these days are, in order: EMF, UML2, UML2 Tools.

OK, no big surprise there. But what's really surprising is that people are still downloading zips. Isn't it easier to get your bits via Update? Isn't that why we provide you, the community, with over 20 Modeling update sites (three per project), so you can pick and choose what features to install, rather than downloading & unpacking zips? (Not to mention the countless other sites provided by DSDP, WTP, BIRT, DLTK, DTP, Mylyn, PDT, Equinox, Eclipse, GEF...)

Hell, with the advent of Eclipse 3.4, you can now even update the platform itself, so the Eclipse experience need only be one zip + a ton of multi-threaded updates. Say what you will about p2 (and I've said my share) but isn't it faster and easier than having to download zips?

I can understand being stuck on a platform version from a while ago, but with Eclipse 3.5 only 4 months away (yikes!), and e4 due next year, should anyone still be using Eclipse 3.0 or 3.1 and EMF 2.0 or 2.1? You're missing out on over three years of new development!

What's even more bizarre to me is that when maintenance releases are available, people still download the point-zero releases. Are we wasting our time doing maintenance? Should we be focusing on the next release instead? (In related news, Ganymede SR2 is just around the corner. Is everyone ready for what will be the final chapter for the Eclipse 3.4 stack?)

Of course the one glaring hole in all this data are the numbers around how many update jars are downloaded. If you'd like to see this resolved, have a look at bugs 239668 and 187968 and vote accordingly. It's really depressing to think that since September 15 we've only had 150,000 downloads (100,000 unique) when in 2006 we broke 1,000,000 downloads for 2006 on December 1 of that year, averaging just under 100,000 downloads a month.

Either the popularity of EMF - and of Modeling - has waned by two thirds, or we need a better way to capture stats.

50,000 Zips

Since turning on Google Analytics for tracking Modeling project zips on September 15, we've gotten over 50,000 pageviews (33,000 unique) of the download tracker. While this doesn't necessarily mean every one of those hits resulted in a download, it certainly shows a lot of interest in downloading EMF, UML2, UML2 Tools, GMF, Query, OCL, and XSD, to name but a few of the more than two dozen Modeling Project components.

Unfortunately, this only scratches the surface -- most of our downloads come from EPP bundles and update sites. But thanks to bug 239668, we will one day be able to track downloads done via p2. If download stats interest you, please vote for bug 239668.

Toward Standardized Groups

June 5 marked the death knell for my conquest for world domination through Eclipse committer groups.

Before then, I was in a whopping 62 different linux groups on dev.eclipse.org, coming in third behind Kim and Sonia (80 and 79, respectively).

However, with Karl and Matt actively cleaning up these duplicate and often empty groups & standardizing the names, my numbers have dwindled to a mere 55. Sadly, I'll lose 4 more when we standardize GEF...

Still, consistency is better than world domination, right?

If your project needs some housecleaning, Karl's looking for volunteers to work with him on this effort. If you are interested, I'd recommend starting with a shared spreadsheet (eg., [1], [2]) to document your before and after states.

Lies, damned lies, and a side order of acronym soup

By my calculations, EMF has scored over 2 million download requests since 2006-01-01. That's a huge number, but it's also an inaccurate number because our stats gen was broken for a while so there's a big hole in the Update Manager download stats. As such it's probably much higher, given the traffic we've had since the stats tool was fixed.

Looking at the overall numbers for all the tracked Modeling projects, we've had 5 million download requests since 2006-01-01 for EMF + M2T + MDT ... and that's just the Update Manager requests!

It's also fun to see who's downloading our bits. As expected, the US tops the chart, with Canada a respectable 8th. To get a sense for where EMF is used consider where the following projects are developed:

1. US: GMF, M2T-JET, WTP, TPTP, VE, ...
3. China: MDT-EODM, TPTP
5. France: EMFT, MDDi, M2M, GMT
8. Canada: EMF & EMF-QTV, MDT (UML2 & OCL), M2T-JET, GEF, GMF, WTP, TPTP, ...
12. The Netherlands: EMFT-Teneo
17. Czech Republic: GMF, MDT-UML2 Tools

To paraphrase Ed and the Zulu people, "umuntu ngumuntu MDD!" ;-)