Actually, I already know what chars windows does/doesn't allow in filenames. Your assumption is slightly wrong though, vis.

C:\test>echo .>'

C:\test>dir
 Volume in drive C has no label.
 Volume Serial Number is 8C87-E163

 Directory of C:\test

02/08/20  02:55a        <DIR>          .
02/08/20  02:55a        <DIR>          ..
02/08/20  02:55a                     3 '         <<<<<!
02/08/13  03:12p                   210 178232.pl
[download]

In an earlier thread the following:

...
@ARGV = map glob, @ARGV;
...
[download]

was advised against because "someone might pass in an escaped (eg.what\*ever) filename containing glob chars", and that might cause the above snippet to "break" on *nix environments. I thought then and I am asking now:

1) Why would anyone wish to embed "*" in a filename? Other than shear perversity.

2) How would this break?

Surely, globbing filespec of "what\*ever" is going to match the file named "what*ever"? Granted it might also match "whatifever" etc. but (see Q.1).

I guess I accept that OS's in general (*nix included) have to reserve some chars in filenames and other system entities (try creating a file "feet/sec" on *nix?).

Right "'" is allowed, D'oh. TANJ. But there are still others. Who are we to say what the user should do? :-P But no, seriously "?" would probably be used quite often, what with the apparent trend of end users to put the file's contents in its name. "*" in filenames might be used to either draw attention to the file, or to make them harder to delete. If you type rmv -rf what\*ver you don't exactly expect whatever to go bye-bye...

--
perl -pew "s/\b;([mnst])/'$1/g"