World's 103rd Most Popular Open Source Package: Woodstox! (with 3% "market share")
(and incidentally also World's 104th and 125th most popular OS package it seems...)
Ok, so the first thing you are thinking after hearing this press-stopping news must be: what the heck are you babbling about? Ok, here it is: Open Source Census (page 3). It is official now, all we need is Netcraft confirming it (hey, can I get my geek badge back for this nerdy inside joke?)
I think this site is interesting, as yet another tangent on how open source "eco-system" (yes yes, it's a buzzword... but for once, fitting) is evolving. Since there is quite a bit of data to play with, well, some folks are playing with it. That's nice: data may not be groundbreaking, but it is still at least mildly interesting. And something that just would not exist for proprietary systems, not for free, and probably not even for fee.
As to how it all works, I suspect the tool that scans machines to produce submissions on which reports are based on correlates names of found jar files with something like info from Maven repositories, or perhaps OpenLogic's certified library database (description on the page is somewhat vague and mostly non-technical -- which probably makes sense I'm sure there are services being sold, as part of the whole experience). Either way, I hope detection logic can be improved: as is, there are no less (and perhaps more?) than 4 instances of Woodstox (part of this is due to 2 alternate licenses, but naming change for jars between versions 1 and 2 contributes to the problem).
Actual rankings are of course interesting to some degree. You can deduce many things which are sort of obvious: for example, that there is lots of inertia in software deployments (ancient versions being used), that most people just deploy whatever defaults system come equipped with (why else would anyone use, say, Stax reference implementation for anything?), and that software stacks largely determine which other libraries get used. Woodstox would not get even its humble 3% coverage if it was not for stacks like XFire and Axis2 (and many other stacks that do use it as their Stax implementation of choice).
It is also interesting to note that of other Open Source packages I have worked with, the other one listed is Java Uuid Generator. (with 1.3% prevalence, which while somewhat low is still quite high considering it is not bundled with many stacks [AFAIK]). And Jackson and StaxMate should make it to the list in future. I especially hope that No Coder Has To Suffer from Writing Pure Stax API Code. Ever. There is no need, now that StaxMate works on all known good Stax implementations. And Jackson: well, considering that it is more than 2x as fast as Woodstox on parsing (when using equivalent document information), what is not to like?
On that note, I better go and finish up StaxMate 1.3. It will be good. You will like it, and will want to use it. Stay tuned!