Having setup LANs for many years and understand why you dont allow 2 DHCP host on the same LAN, I recently ran across 2 solutions where I need to appeal to you guys/girls for an explanation.
I first discovered this in Knoppix, 2 years ago and recently found a similar model in Puppy. In Knoppix it is, incorrect in my opinion, referred to as Terminal Server and in Puppy it is called Netboot...but they do the same thing for their LANs. They will provide everything needed so that another PC can use its BIOS to direct the PC so that if a "PXE" host is listening on the LAN, it will get its boot image and boot from the LAN instead of local storage devices. This ability works from a LiveCD running either Knoppix or Puppy.
Scenario: I have a router with LAN PCs all of which get their IP creditials from the router's DHCP service. One of the PC is Knoppix for example (or, it could be Puppy. It doesnt matter because the discussion is the same).
Question: Now for my request for explanation. The PXE host (Knoppix/Puppy) starts it service to allow the LAN PC to connect to boot. That service includes a DHCP service being started on the PXE host. When the LAN PC starts, it broadcasts for DHCP credentials.
How does that LAN PC know to use the PXE host's DHCP to boot, then use the router's DHCP once its OS starts the console desktop? Can anyone explain this?
Additional Info: My Puppy logs show that the DHCP address range that he is using is the a different range from my router's range. Both DHCP host are operation on the same LAN with the same subnet. Still, how does the LAN PC know which to use??? Is a PXE request coming from a LAN PC somehow different from a "normal" DHCP request for service? Is this why during PXE operations to get the LAN PC to boot, the LAN PC has an IP in the PXE host's range, then, when the LAN PC boots, how does it know to use the IP from the router's DHCP service? What constitutes/causes this magic?