I wouldn't suggest that using delays is a 'bad' thing as the data sheet tells us the execution times for the various instructions - add some fat and we will comply with worst case timing. Nothing bad at all. After init the two things I do to the lcd is to move the cursor and write characters and normally using the 4 bit interface. Obviously polling for the ready bit is more time efficient but by doing a write only interface I save 1 port bit. For most cases saving a port bit is worth way more than a little bit of execution efficiency.