So, how do you know what key1 is a representation for. 
To put it another way... How do you know what column value key1 is a u
+nique value in? Or am I missing something obvious?

I'm trying to get a picture of this dataset built off of a database. D
+o you have a simple snapshot of the datasets you might be receiving?
[download]

Not the real data but think of it as

apple      red
orange     orange
grape      green
apple      green

Neither the keys nor values must be unique; the script will take care of merging the multiple values for a single key, if needed. The source database thinks it is outputting key{tab}value{newline} but, because of the different line-endings, I get key{tab}value{\cM} instead. At least, I do when the end-user creates the file on a Mac. But, other end users can create this source file on a Windows system where I get different line endings so I need to accommodate any line ending. This is also one of three source files which could all come from different sources before the end-user uploads them using this CGI script.

I was hoping there was a Perl/CGI.pm equivalent of FTP's "ASCII" mode.

Sounds like a tough spot ur in. The only thing I can suggest
 is to either get the user uploading the data to do it in a 
specified format or do your best to try to find key-to-value pair patt
+erns.

If I was in the spot ur in and couldn't find the carriage 
return value possibilities, then I would probably try to 
pursue the infile formatting standards as best as possible.
Things like # of columns in the table or requiring the user to 
make the first line a 'header record' so you could see where
the key value pairs repeat.

Best of luck
[download]