-	Auto-negotiates (although 'negotiate' is a misnomer, as we have discussed) speed & duplex
    o	Caveat:  while the standard supports half-duplex . in the field, you'll only ever see half-duplex in 10 & 100 . no one ever implemented a chip which did 1000/half . and above that rate, the standards do not even include half-duplex
    o	Thus, the complete list of what you'll see in the field is: 10,000/full, 1000/full, 100/full, 100/half, 10/full, 10/half.

SFP Ports
-	Accept SFPs (10M, 100M, 1G)
-	The port auto-detects the speed of the transceiver
-	No auto-negotiation.

SFP+ Ports
-	Accepts both SFP as well as SFP+ (typically 10G, more recently 25G . 2s-f-esx supports 25G).
    o	The list you'll see in the field:  25G, 10G, 1G, 100M (rare)
-	On some platforms, the port auto-detects the speed of the transceiver, on others it does not and one must specify the speed, e.g. "interface Eth1/12; speed 10000".  We have seen inconsistency in this within a single switch, i.e. some ports requiring the 'speed xxxxxx' setting and others not.
-	Everything is full-duplex; half-duplex does not exist
-	Most SFP+ ports I've seen service twinax transceivers
-       Notice that you will never see a 10GBaseT SFP+.  Why not?  10GBaseT is power-hungry, and the SFP+ specification does not deliver enough power.

QSFP+ Ports
-	Accept QSFP+ transceivers, the only ones I know about support 40G or 100G glass transceivers
-	The port auto-detects the speed of the transceiver
    o	Caveat:  you can typically insert an SFP+ adapter, which then allows you to then insert an SFP+ transceiver . you typically then have to hard-code speed (e.g. "interface Eth1/12; speed 10000"), and now you run your expensive QSFP+ port at 10G
    o	Be aware that some hardware platforms will then downgrade all their QSFP+ ports to 10G at this point - you may not be able to mix & match 40G and 10G within a cluster of QSFP ports