I'd actually checked out your module already. :) It looks like a nifty solution to a vexing problem.

For Kinosearch, I'm content to have systems which can't deal with large files run up against the file size limit. If you're running a search engine where the index exceeds 2 Gb, you're running a serious search engine, and you're probably not running it on a machine that can't deal with >2 Gb files. My main reason for writing the workarounds was a perhaps mistaken impression as to what maximum file size 32-bit Perl's internal file manipulation routines can handle safely.

WRT to problem of auto-growing a file, I was doing it by using a scalar write buffer. That stopped me from having to check if I was exceeding the 2Gb limit every single call to print(). Perhaps a similar solution could work for your module?

-- Marvin Humphrey