I'm curious though, CGI->upload() returns a filehandle, so does CGI->param() for that matter. I am wondering why in _handle_file() you read from this filehandle and write to a new tempfile so you can provide your own filehandle. Wouldn't it be easier to use the filehandle the CGI.pm provides?

I have actually considered this but haven't had a chance to read through the CGI.pm in detail to see exactly how it handles its temporary files. Much of this module code has evolved from subroutines that I have used in some CGI scripts in the past.

If anything, you might view the method I have taken as the defensive approach in the absence of any better informed opinion :-) This can also be seen in my resetting of the file pointer for the temporary file after calling the checktype_filehandle function from File::MMagic - While this resetting of the file pointer is in all likelihood unnecessary, I have included it for the sake of a "defensive" approach to this code.

As for the binmode on the newly-created file handle, this is a very good point - Again, I will look into this and the handling of temporary files by CGI.pm and revise this code accordingly.

Thanks again for you comments $code or die++

Update - I've had a look at the implementation of temporary files within CGI.pm and at this stage, I believe I will remain with the implementation I have in the code above - This is in part because of the robustness of the POSIX-based new_tmpfile method exported from IO::File which I am quite comfortable with from a "paranoia" perspective :-)

perl -e 's&&rob@cowsnet.com.au&&&split/[@.]/&&s&.com.&_&&&print'