The PXE daemon allows network managers to boot PXE-enabled machines over the network. Basically it envolves BOOTP, TFTP (so far so good) and a PXE daemon. This is where all the trouble starts. There is only one implementation arround, and if doesn't work on anything other than little endian machines.
I originally tried to correct some of the major problems in this code. However firstly, I could not find anyone with at Red Hat or Intel to deal with and secondly, Intel seem very unintrested in porting the code the big endian machines. The source is also very messy and it would have taken longer to correct the original code, than to re-write the whole daemon. If you don't believe me on this point, just have a look at it, IMHO any daemon that responds to keyboard input is a bad thing (it thinks it had packets coming from the keyboard even when in daemon mode)
You are lucky then, but a lot of places have already got a bootserver, and most likely it will be a UNIX based machine that is not x86. So, why put in another machine when there is already one there?
If you know enough about PC, you will know there are several operating modes. The mode the PC boots in is only able to access the low 640KB of memory. By using the PXELINUX bootstrap code it is possible to boot using the full amount to memory in the system. The means you can do some nice things including boot a Linux kernel + Ramdisk over the network, as you would for a Sun. Please note that the Intel bootstraping code is very limited in this sense.
I dont recommend it anymore, I just havent had time to do a lot of updates, PXELINUX is far far better (see link below). However Intel's PXE daemon is naff (IMHO)
What's New in This Release:
· Increase the interface buffer to hold more than three interfaces
· Fixed a segfault in when multicast was disabled
· Fixed null pointer dereference when looking at interfaces with no IP address