store & forward = entire frame received, held, CRC'd & then forwarded
cut through = frame forwarded out the appropriate port as soon as the destination is determined
fragment free = the first 64 bytes of a frame are checked (collision window) and if all is good the frame is forwarded.
Catalyst1900 and (I believe) the 2900 series that is replacing them come with fragment-free enabled by default.
These are also the most common cisco switches out there, whether singly or in a fabric.
And most if not all cisco switches give you packet switching options.
I'm not putting my hand on a bible about this because I'm not as certain about product details as a salesman would be and frankly, I'm still amazed that I managed to pass the CCDA.
But...
store/forward has the highest level of latency of any method.
These days the transport layer does the error correction heavy lifting and you can get good QoS by just insuring the first 64 bytes aren't screwed up.
And, I believe that the original question was about what I liked and didn't like.
I didn't realize that my opinions were going to be a problem.
That's why I'm bowing out of this discussion.
I have other forums that satisfy my need for flame wars.