Do all Ethernet Frames have a Src MAC and Dst MAC : No router involved

[Cisco10956]

Hello

  Will an Ethernet Frame always have a Src and Dst MAC address of the destination computers network card. 

Travel of a frame on a switch.

Ethernet frame = [ "IP Packet " : Src:MAC : A Dst:MAC_B ]

[Jon Marshall · Accepted solution]

Yes they should do although it won't always be the address of destination host ie. it could be a broadcast or mulitcast address. 

I suspect when you have read that the destination mac address is unknown it doesn't mean there isn't a destination mac address, it means it is unknown to the switch ie. it is not in the switches mac address table. 

Jon

[Joseph W. Doherty]

"I suspect when you have read that the destination mac address is unknown it doesn't mean there isn't a destination mac address, it means it is unknown to the switch ie. it is not in the switches mac address table. "

Ah, only @Cisco10956 could clarify "unknown" (ideally with a reference to what he has been reading), but since he keeps harping on switches, although discussing host-to-host communication, that seems very possible.

[Leo Laohoo]

Yes and no. 

As a rule, "officially", every device must have. 

However, I have seen devices "in the wild" that has no MAC address and this is why they have to be hard-coded with static IP addresses.  And these devices are used in "health care" environment and endorsed by the US FDA.  

Go figure.  

[Joseph W. Doherty]

@Leo Laohoo that's interesting!

Do you have any references to these devices that explain how they communicate?

I can envision something like the sending host creating a static ARP entry, for the special destination device's static IP (where the [unicast] MAC doesn't matter [beyond not being used by a non-special host]) and where the special destination device, basically, runs in promiscuous mode (so it doesn't really care or match for a unicast MAC).

Hosts, though, not having their own unicast MAC, would seem to force switches to effectively unicast flood all these device packets.

Laugh, does seem fitting for US "health care" as it probably allowed some medical appliance vendor, way back when save some $ building their own custom NIC, which incidentally, is one reason their devices "need to be" priced at 10x, or more, vs. standard hardware doing the same (or more) functions.

[Leo Laohoo]

---

@Joseph W. Doherty wrote:

Do you have any references to these devices that explain how they communicate?

---

¯\_(ツ)_/¯

DALI light controllers is a good example.  

They are hell to troubleshoot and implement because: 

• The manufacturers do not have an idea what a "MAC address" is
• Documentation is a three-dollar-bill.  

The controllers and the clients (10 Mbps, Half duplex) operate in some "funky" broadcast.  They do not talk MAC address and all of them needs to have a static VLAN (because DHCP is not an option to "tick").  

Excuse me, but I really need to barf.