At the risk of stating the obvious, for your 'lines 1034-1047' problem I can see two efficient solutions.

1: If you have control of the source file make it fixed-width (possibly not even newline terminated records), then seeking a particular 'line' (record) is a simple bit of math and a seek().

2: If the file must have variable length lines, and assuming you do the mid-file extraction frequently and that the file persists for a relatively long time...
Scan the file first for newlines.
Record the position of each newline in a second index file.

To look at a particular line:
Open up the index file.
Scan to the line# you're looking for.
seek() on the file you want to look at.

If the requested lines are past the end of the index:
seek() to the last known newline in the source file
index to the end (appending to your index file).

Another optimization is to store the index in binary format (ie: integers) so that you can find the index for line #27231 simply by seeking to (size_of_rec * 27231).

The algorithm in 2 can be implemented in Perl, but would be simple to do in C or C++ which would make it run pretty fast too.

Steve-