[Xlock-develop] [Fwd: Re: [Xlock-discuss] PAM integration]

[Yuri Bushmelev]

On Tue, 07 Nov 2006 14:34:58 +0100
Dan Lukes <dan at obluda.cz> wrote:

> > Ok, I have new auth framework (draft) instead of one in TODO:
> > 
> > Xlock will be started setuid root. After start and proper
> > initialization we create two pipes (or socketpair, if available) and
> > fork a child (auth-backend). Parent (main xlock process) will revoke
> > all privileges.
> 
> 	It seems you are planning "another PAM" which act almost as PAM, but 
> has own modules - with legacy PAM as one of them.

Yes. It looks like "another PAM". But while we support some platforms that
does not have PAM, we must support some internal methods..

> 	The "minimal privilege concept" is good, but I'm not sure we need the 
> separate process and IPC protocol between them really. It seems to be 
> over-complicated to me. The careful use of seteuid() can satisfy my 
> paranoia.
>
> 	Well, complete separation is more secure, so if you can spend the time 
> for it, let's go.

I'm not sure now.. When we properly sanitise command line parameters
and user input (passwords for internal methods), complete separation is
really over-complicated.

> > AUTH\r\n<username>|<uid>
> 
> UID is not unique identification of user so it can't be used there.

Accepted :)

> > I have question too.. Such system is not too compex for xlock?
> 
> 	IMHO, yes, it is.
> 
> 	But the decision is up to David, as he is project leader and up to you 
> as your time will be spent.

Ok, let's wait for David's anwer.

> P.S. I sure I will understand even you wrote text it in russian, but I'm 
> worry about David ;-)

My native language is Russian. In scool and hi-scool I have learn
German. Now I write English. What is next? ;)

-- 
Yuri Bushmelev