> It's fine for writing (as shown of one of those pages) but I found it
> troublesome for reading.  Basically you'd have to base generic
> readline, gets and such, on a class-specific implementation of
> something as small as a getc.

Why? I would suggest that the implementation relies on read(byteCount).
You can mantain a buffer (e.g. with gets if you can't pass all of it on
to the caller).

> Can you see yourself iterating through
> each byte until you find the pattern?  I _think_ it's a bit
> inefficient compared to (at least non-greedy) regexps.  (I'd be happy
> to be proven wrong, though!)

Yep. getc would be too inefficient, but it is not necessary.

Thomas