On Thu, Jan 24, 2002 at 05:01:39AM -0500, Sandy Harris wrote:
> Thomas Steffen wrote:
> IPsec provides encryption and authentication at the IP layer, protecting
> everything running above it. If your "X terminal" runs Linux, get the
> IPsec from www.freeswan.org. I believe all the *BSDs now include IPsec,
> and Sun and HP offer it for their Unices, though I'm not sure if it is
> built in or an optional extra on those.

I think IPsec is my long term answer, even though the vendors appear to
be dragging their feet about adopting it.

I'm not concerned at all about performance loss.  ssh is more than
liveable in that regard.

The current X is subject to MITM attacks, but I'm vastly more concerned
about replay attacks.  And xhost +, but that's another story.

> > > I'm not that worried about MITM attacks,
> 
> Methinks you should be. 

Whatever.

> > > The other guy had it right - it has to be a no brainer, and work on
> > > a true X terminal.
> 
> It's going to require some sort of software upgrade to the terminal,
> whether you put crypto into the X libs, use SSH or use IPsec. That
> means you need either co-operation from a vendor or an X terminal
> with an open development environment. Do you have either?

My hope isn't to come up with an immediate solution for my own
applications - I have that, by just avoiding X terminals and using ssh.
My hope is that my support group could realistically start supporting X
terminals again in the long term, without feeling like we're encouraging
our endusers to do the Wrong Thing.  If that meant writing some
opensource stuff to help start a needed standard, that could happen.

> > What about getting ssh included into the X terminal? I guess there are
> > problem with xdm, but appart from that I cannot see what you could
> > gain by including encryption in X.
> 
> Nor I. To do any network crypto well enough that people can rely on it
> you need both well-analysed crypto primitives (lots of those around)
> and a well-analysed cryptographic network protocol (a big, tricky
> undetaking).

Mostly agreed.

xdm seems a crucial bit though.

> SSH meets those criteria, is easily supported on the host end, might
> not be too hard on the terminal end, and is adequate for this task.
> It is the obvious simple solution.

This has become academic, but I would think ssh would be kind of a mess
to integrate.  My expectation is that the authors would say
"Modularization?  We never planned on anyone needing it."  But hopefully
I'm wrong about that.  Sometimes modularization does pop up in
surprising places.

> IPsec also meets the criteria and is more general and flexible than
> SSH. However, it is a bigger implementation project and it is not
> clear that the generality pays off in this application.

To reiterate: I'm content with this.  I guess.

> SSL/TLS also meets the criteria. I don't see offhand how it could
> be applied to this problem, but there is probably a way.

Sure.

> Trying to build crypto into X strikes me as re-inventing the wheel,
> with a risk of having it come out square.

Naturally.  Of course X's present security is more of a triangle.

Thanks to all respondents.  I think I'm going to let the thread die and
unsubscribe.

-- 
Dan Stromberg                                               UCI/NACS/DCS

Attachment: msg03530/pgp00000.pgp
Description: PGP signature

Reply via email to