We had an interesting exercise at the recent Madison Perl Mongers meeting, where we took an ugly Windows shell script and turned it into a Perl script. The resulting Perl was noticeably longer than the original script, but also had functionality broken up between subroutines, and was far easier to understand.
One of the sub-problems was to get the current free space on the disk. After some searching with Google and Perl Monks, I found Win32::DriveInfo. Problem is, it isn't cross-platform.
I didn't want to bring up a counterargument right then and have to explain it out, because we could have been debating it the rest of the night. But here's my position: I don't care if it's cross-platform or not. But I don't care in a carefully considered way.
This particular script was for an automated daily sysadmin task on a couple of Windows servers. There is no guarantee that it will ever need to run on anything else. So what I would do in the real-world is make a short search for a cross-platform alternative. If nothing came up within a few minutes, I would shrug my shoulders and wrap all Win32::DriveInfo access into its own subroutine and move on.
As it happened, some people already knew about a cross platform solution (I believe it was Filesys::DfPortable), which is fine. But if I hadn't found that (and it's really not a good, easily searchable name for a diskspace module), I wouldn't worry about it.
It might be that I never have to port it to another system. In that case, my search would be a big waste of time. What I need to do is write it in such a way that it's easy to pull out the old behavior and substitute it. This way, if I ever do need to port it, the job will be easy.
Furthermore, you'll almost certainly make hidden assumptions that aren't cross platform, anyway. The history of Java ought to be a lesson here.
One of the goals for this committee hacking session was to make the script cross platform, which was fine. I just don't see a reason to get overly concerned about it, provided you otherwise write code in a modular way.