Window size and minimum frame size



just read the book of chris sanders “practical packet analysis” and just finished the chapter on slow network issues.
there he mentions an example of tcp windowing.
and he writes that if the client and the server agreed on a window size of 5000bytes the client sends one time 2500bytes and then 2000bytes and then the server would respond with a ACK package. but i thought the server would only acknowledge the client if the window size is reached, or not?
because according to the book the client knows about the window size and thus could adjust to it, or not?

and furthermore i see in some sample captures from the book that e.g. an arp frame is only 52bytes big, but i thought the minimum frame size of ethernet is 64.

any ideas are appreciated.