LDM_DIRECTX=yes by default?

Warren Togami wtogami at redhat.com
Mon Jul 7 02:54:23 UTC 2008

Robert Arkiletian wrote:
> On Thu, Jul 3, 2008 at 10:16 AM, Warren Togami <wtogami at redhat.com> wrote:
> ...
>> This is bad for security, but if our goal is to have something usable
>> out-of-the-box in a similar fashion to how K12LTSP was, then perhaps we
>> should do it?
>> How do people feel about this?
> Not sure how one would exploit this security hole? So traffic is not
> encrypted. It travels from the client to the switch to the server. How
> is someone on another client or even with a laptop on the lan going to
> sniff keystrokes? If they fake the MAC address of a client that X
> session will break anyway. Unless one is root on the server and
> captures traffic with wireshark on the internal nic I can't see how to
> spy on the traffic.
> With ldm_directx=yes
> My only concern is if I can safely su to root from a client without
> having to worry about some clever kid sniffing my root password.
> If this is not safe then please enlighten me as to the exploit method
> as security through obscurity is no security.

This isn't really security through obscurity.  It is known to be 
completely wide open and unencrypted on the wire.  This is no worse than 
how LTSP worked in past years.  Except with LDM at least your initial 
login and password are encrypted.

If you want to make it unencrypted for ordinary clients but encrypted 
for specific clients, you could use MAC addresses in lts.conf to control 
LDM_DIRECTX for those specific clients.

Warren Togami
wtogami at redhat.com

More information about the K12Linux-devel-list mailing list