Three options for you:

1. Build an index of line numbers to offsets.
2. Use DB_File's DB_RECNO interface to tie the text file to a hash.
3. Read through the file each time.

Indexing is easy (code untested, as usual. You can find tested similar code in the Perl Cookbook):

open(IDX, "> index") or die;
open(DAT, "< datafile") or die;
$pos = 0;
do {
  printf IDX "%8d" $pos;
  $line = <DAT>;
  $pos = tell(DAT);
} while (defined $line);
close(IDX);
close(DAT);
[download]

and to use it:

open(IDX, "< index") or die;
open(DAT, "< datafile") or die;
# assume $lnum is the line number to go to
seek(IDX, 8*$lnum, 0) or die;
read(IDX, $offset, 8) or die;
seek(DAT, $offset, 0) or die;
# now DAT filehandle is positioned to read the
# record you wanted
[download]

The DB_RECNO interface from DB_File is certainly easier to use, but it's probably not as efficient as it can't keep the index on disk as my code here does.

Hope this gives you some pointers.

Nat