I assume you have RHEL3, given that version of Perl. There is a bug in that Perl that shows up when reading data from a file. It's releated to unicode. RHEL3's default locale is 'lang.UTF8'. If you set the LANG to 'C' before running your script, it works around that particular problem. You can't do $ENV{LANG}='C'; in your script; by then it's too late. You have to do something like this (example in bash):

> export LANG=C
> perl myscript.pl
[download]

So you could write a bash wrapper that did that as a workaround.

Unfortunately, bugzilla.redhat.com is down at the moment or I'd chase down the entry for you. The last I heard, it was supposed to be fixed in update 3. I have just confirmed that that is the problem here.

If you can't get by with a bash wrapper, I suggest that you build Perl yourself from source.