I have a need to do a couple of on-line surveys and to that end I've been installing and trying out some open source survey packages.
The ones I've tried have all worked, but they've all had something in common with lots of other open source products I've downloaded and installed - they have a show stopper bug you crash into during installation or startup that's relatively easy to fix, but frustrating to run into.
Mod_survey, for example, was both one of the most powerful I tried and one of the most frustrating because some of the installation instructions are simply wrong -so if you don't happen to know how apache configuration works you're going to spend a long time trying to figure out why the include statement suggested in the documentation isn't the right way to go.
phpESP, in many ways my prefered choice, doesn't actually do the one question type I really need, but it too had a trivial "techie-enough?" test in the installation procedure.
I settled on phpsurveyor: it's great and does just about everything you would hope for except that the installation is so automated I spent a confused twenty minutes figuring out that it's kinda like a Unix process: you just start to use it, and if the context is in place, off you go.
I've seen this kind of thing a lot - compile procedures that die on spelling errors in include file names, install instructions that have obvious mistakes, makefiles that force bad library links, and so on: all of them things that couldn't possibly work in any test environment anywhere.
So how do they get into the downloadable tar.gz files? My theory has long been that they're intended as a kind of hoser test: if you're too dumb (or uneducated) to see past them, you don't deserve to run the package.
If that's what's going on, I kinda think I agree - except for one problem: if open source is truly going to succeed, we've got to make it work for the hosers too.