> > Just configure "pxelinux" from "syslinux" (use freshmeat.net to find it) > > to boot the kernel and the new initrd. > > Exactly: you use pxelinux *instead* of Etherboot, which is perfectly OK, > but contradicts your previous statement.
Ok. You are right. I just "ignored" the fact that "pxelinux" is also a bootloader. > Whether you choose pxelinux or Etherboot is a matter of taste: it is > fairly easy to change to one from the other; at the moment both have > their advantages, but it seems clear to me that Etherboot has more of a > future for non-trivial network configurations. Just one question from the non-Etherboot expert: Could you load a kernel and an initrd? The network-card I have (Intel i82562, built-in) has severe problems with Linux' eepro100 driver in 10MBit/half duplex: It just hangs the machine. The problem is supposed to be fixed in linux-2.4.11 or higher, but ever since I got hold of Intel's drivers (e100), I use them as they worked ok in all situations. BUT: The e100 could only be built as a module, so I heavily depend on the "initrd" to load that driver. > BTW - there's nothing wrong with H. Peter Anvin's pxelinux. Though the > existing Etherboot/PXE code doesn't include any of it, I know that Vasil > did look at the pxelinux assembler (yes, assembler, not C). My main > criticism is that pxelinux shows its origins from the syslinux disk > loader; it is not really a starting point to develop interesting network > booting behaviour. You are right. In a first stage, I wanted to remove the "TSIZE" requirement from the code, without any success... > Currently one real advantage of pxelinux is its use of UNDI, the PXE > device independent network card abstraction. I agree totally. > The existing Etherboot does not use UNDI, so that you have to ask rom-o-matic > for the right nic, and it's possible that pxelinux may work on a PXE NIC which > Etherboot does not support. How about the network drivers. They could also have bugs (see above) which prevent Etherboot to function even if the driver is the correct one for the chip. I think the UNDI (in "BIOS") could be fixed by the vendors... > Anyone know of an example, just out of interest? It would be nice to port > Etherboot to UNDI to give it the same power. Such a port would probably use > pxelinux code. Anyone looking for a useful assembler project? Yes, anyone? > pxelinux.cfg instead of adding lines to dhcp.conf, and don't mknbi your > kernel. If this works for you, fine. If I had a large network with many > different network card types all with PXE nics, I'd use pxelinux too. > I do not wish to discourage anyone from trying pxelinux. I just wish to > correct the assertion that PXE, or a boot rom PXE client somehow does > work instead of (or "better than" or "simpler than") Etherboot. Not so. That was not my intention. I just stumbled over the "non-UNDI" issue.. So long, Jürgen ---------------------------------------------------------------- Jürgen Zimmermann Department of Computer Science [EMAIL PROTECTED] University of Kaiserslautern Phone: +49 631 205 3280 P.O. Box 3049 Fax: +49 631 205 3299 67653 Kaiserslautern, Germany WWW: http://wwwdbis.informatik.uni-kl.de/staff/Zimmermann/ ---------------------------------------------------------------- "The Number of UNIX installations has grown to 10, with more expected." -- The UNIX Programmer's Manual, 2nd Edition, June, 1972 All systems working within normal parameters - BSD: OK, Unixware: OK, Linux: OK, NT: Critical, Win98: Crashed _____________________________________________________________________ Ltsp-discuss mailing list. To un-subscribe, or change prefs, goto: https://lists.sourceforge.net/lists/listinfo/ltsp-discuss For additional LTSP help, try #ltsp channel on irc.openprojects.net
