On 2011 Oct 04 (Tue) at 12:12:38 -0600 (-0600), Theo de Raadt wrote:
> > On 2011 Oct 04 (Tue) at 12:00:08 -0600 (-0600), Theo de Raadt wrote:
> > > > note the definition of def_anon in mountd.c, that one has a uid/gid of 
> > > > -2, now
> > > > this might be a problem I suppose.
> > > > 
> > > > However, for the nobody case. If you are mapping to nobody, mountd will 
> > > > populate
> > > > the exportlist correctly for the anonymous user (with the nobody 
> > > > information)
> > > > and pass that to the kernel, where there is some messing about...
> > > 
> > > I don't think this changes anything regarding the master.passwd commit.
> > > 
> > > Obviously nobody isn't just for NFS.  It is used by locate, too.
> > I dont know what locate is doing with nobody.
> > 
> > What seems to be lost here, is that up to a point the nobody and NFS is
> > analogous to _daemon/daemon. 
> > 
> > So, thats the point of nobody and thats why there is a linkage between
> > NFS and nobody.
> 
> I think you've not read the diff that was commited.
You'd be wrong.

> master.passwd used to say:
> 
>     "Unprivileged user for NFS".
Which I'm arguing it is. Just like _bgpd is the BGP daemon user.

> But locate uses that login, too.  If it wanted to be very
> truthful, it could say
> 
>     "Unprivileged user for NFS and locate(1)".
Again, I have no clue what locate(1) is doing, but given this I
suspect it is wrong and warrants some looking into.

> Or it can just say
> 
>     "Unprivileged user".
> 
> That is what was commited.
And this I think is wrong. There should be no general unprivileged user.

There should be ,,application specific'' unprivileged users.

Am I the only one who sees this parallel between the numerous _daemond
users and there respective daemons and nobody and NFS ?

If so, can someone please enlighten me about the purpose of nobody? Is it
going to be used (do we want to encourage the usage that is) for something
else or new ?

Reply via email to