Am Montag, 7. Juni 2004 11:59 schrieb John Sutton:
> Hi there
> To date I have been using netboot's mknbi (version 0.8, just because it was
> easier to compile than etherboot's mknbi back in 1998 and I haven't seen
> any need to change it since ;-) while at the same time using etherboot's
> rom code dowloaded from romomatic.net as and when I get new cards.  Anyway,
> it seems like the world has been passing me by as it now appears that
> etherboots nbi code is capable of booting up a kernel image which has got
> an initrd strapped to the end of it?  Hence it is no longer necessary to
> build all the network drivers into the kernel, which is a big plus for

Using more recent etherboot version, you'll need a recent mknbi as well. You 
have been warned :-)
There is a lot more possible with mknbi & etherboot, like booting DOS etc...

> abandoning my own config. But this extract from the docs has confused me:
>
> ---------------------------------------------
> LTSP - Linux Terminal Server Project - v3.0
>
> Chapter 1. Theory of operation
>
> 16.  dhclient will then be run, to make another query from the DHCP server.
> We need to do this separate user-space query, because if we depend on the
> query that comes from Etherboot, it will get swallowed up by the kernel.
> And, the kernel will ignore any NFS server that might have been specified
> in the root-path. This is important if you want the NFS server to be
> different from the TFTP server.
> --------------------------------------------
>
> Surely all the info from the initial dhcp query is available on the kernel
> command line so why is it necessary to do a second one?

I wouldn't expect all initial info to be available on cmdline. This will 
especially be true in setups where the ltsp kernel + initrd are started from 
harddrive instead of network or from CD or whatever. Running dhclient makes 
the initrd much more generally usable.

> Finally, in my own config I mount the servers root as root for the clients
> as well and then mount private /etc /var /tmp & /dev-state on top for each
> client individually.  Probably this is considered a security no-no?  In my
> own environment it works fine and IMHO reduces considerably the maintenance
> required when upgrading the whole system, but I'd be interested in reading
> any discussions of these security issues with a view to changing my mind!

The first reason for not doing so is that you need a same-architecture server 
in your setup. LTSP is, contrary to this, able to run off a Win NT server 
even, or any Linux/ppc, Sun Solaris, Open/FreeBSD, Linux/ia64 etc. pp 
servers. Only stuff the server must support (can be splitted to different 
servers even) are NFS-server, DHCP-server, X-login manager or RDP server, 
depending on what you want to do.

The second reason is of course security considerations, as having / exported 
over NFS might reveal /etc/passwd etc. over the wire - If you can do without, 
the better. LTSP only needs a separate subtree available with NFS.

The third reason is that LTSP has some non-typic setups (like local-devices 
etc) that can be much easier supported when the system is in a separate 
directory. Updating LTSP is easy as "tar -xzf" anyway, you can usually just 
unpack the new tarballs (remember to save your config before doing so).

HTH

Anselm


-------------------------------------------------------
This SF.Net email is sponsored by the new InstallShield X.
>From Windows to Linux, servers to mobile, InstallShield X is the one
installation-authoring solution that does it all. Learn more and
evaluate today! http://www.installshield.com/Dev2Dev/0504
_____________________________________________________________________
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.freenode.net

Reply via email to