Re: [Qemu-devel] QEMU: VNC

[Fabrice Bellard]

Daniel P. Berrange wrote:
> On Mon, Feb 19, 2007 at 12:41:53PM -0500, Christopher Olsen wrote:
> > On Monday 19 February 2007 12:30, Daniel P. Berrange wrote:
> > > On Mon, Feb 19, 2007 at 03:11:15AM +0100, Johannes Schindelin wrote:
> > > > Hi,
> > > >
> > > > On Sun, 18 Feb 2007, Anthony Liguori wrote:
> > > > > Christopher Olsen wrote:
> > > > > > Sorry I'll attempt to use the preferred patching method in the
> > > > > > future..
> > > > > >
> > > > > > Secure vnc auth method the default built in method from
> > > > > We can't take a password from a command line.  Supporting VNC auth is
> > > > > super easy otherwise.  I really think we need to have a config file
> > > > > before we can do VNC passwords.
> > > > No, you should not do VNC passwords. The default VNC password exchange is
> > > > insecure and you should not lure users into believing in that false
> > > > security.
> > > Sure it is insecure over an unencrypted network channel, but if you are
> > > tunnelling the VNC connection over SSH, or have restricted it to only
> > > bind to 127.0.0.1  then AFAIK it is just fine. So supporting VNC password
> > > auth would allow users on a shared machine to secure the console from
> > > other unprivileged users on the same box. Definitely useful over the
> > > current situation where there's no way to secure even the local-only
> > > case. For a serious general purpose authentication I'd like to see the
> > > TLS protocol extension for VNC (as implemented in VeNCrypt) supported
> > > allowing both secure auth & wire encryption.
> > >
> > > Dan.
> > I've Checked out the VeNCrypt.. Looks a little win32 oriented... 
>
> Guess you missed the 'unix' directory - I have compiled both server & client
> of VeNCrypt on Linux no trouble.
>
> > I'm gathering the problem here is that VNC is spinning off in many 
> > directions...  So any implementation on the QEMU side will of course marry it 
> > to a particular VNC branch or I had an alternative idea..
>
> I think the crux of the matter is that RealVNC sell a commercial version
> of VNC which offers real encryption. So I'm guessing that's why they've
> never merged any of the patches to do TLS encryption in the open source
> codebase. All the patches for VNC + TLS i've seen posted are iterations
> of each other - VeNCrypt is the most complete implenentation of any of
> them, so the one I'd go for out of the all the choices. 

On the technical side, adding OpenSSL support in the current VNC 
implementation is QEMU seems easy (OpenSSL has a non blocking API which 
can be used with the current callback API).

Fabrice.