Case in point: I finally upgraded 5.6.0 to 5.6.1 on my company's Sun server about two months ago after I finally encountered a show-stopping bug: when testing cpanplus, I hit a segmentation violation in LWP::Request. Run it in the debugger and it works fine. Run it from the command line and it crashes and burns. Eeeew, my favourite sort of bug :)
The thing is, everyone said for ages that 5.6.0 really stank more or less from the beginning, and it was rapidly deprecated, and replaced by, 5.6.1. So I knew that I was treading on thin ice, but until I ran into a bug I figured I'd wait. It turns out I was able to wait for nearly two years.
What I can no longer recall, though, is how quickly did people realise that 5.6.0 was a doomed release? How does that timeframe compare to the amount of time that 5.8.0 has been released? In any event, I'll probably install 5.8.0 on a couple of minor servers, just as I'll continue to keep 5.005_03 running on another server. I'd just hate to install 5.8.0 only to find out that I need to then install 5.8.1
And then I guess the next step is to get around finally to installing 5.9.x when they start to appear, somewhere it doesn't hurt too much if things do break (for the good of the community and all that).
update: in this thread tilly mentions that 5.6.0 was considered a bad release. This was eleven months after the fact. Can anyone find an earlier reference?
update 2003-07-02: Speed of 5.8.0 versus 5.6.1 would appear to sum up the issue, about one year after the question was first posed...
print@_{sort keys %_},$/if%_=split//,'= & *a?b:e\f/h^h!j+n,o@o;r$s-t%t#u'