That's an interesting comment, because the operating system doesn't think of a file as a collection of lines either. One of the motivations for making it easy to read files on a line-by-line or record-by-record basis in Perl was to match how people think, but that may rely on what people want to do with Perl more than any gestalt intuition about the nature of files.

As I said, I am not a programmer, so the finer points of programming are lost on me. As an user however, and more as a social scientist, I have to deal with statistical data and/or text documents. For statistical data, especially time series data, I feel more comfortable in having the computer read/process a whole variable (all records) at once rather than piecemeal. For text documents, I also feel more comfortable in having the computer read the entire document at once and then process it according to my requirements. Bottom line however, I am open to any technique that gets the job done with minimum programming effort on my part. But that's just an user's perspective. I appreciate that a programmer's perspective may be very different.