> From [EMAIL PROTECTED] Fri Apr 30 13:45:34 1993
> To: "Richard Basch" <[EMAIL PROTECTED]>
> Cc: [EMAIL PROTECTED]
> Subject: Re: IP acl & system:authuser
> From: Paul Howell <[EMAIL PROTECTED]>
>
>
> "Richard Basch" writes:
> > It is fine to assume that there may be times when this weak level of
> > security is appropriate, such as accessing system-level software prior
> > to a user logging in, or for machines to save statistics in a central
> > repository when no authentication is possible. However, this should be
> > explicit decisions and system:authuser should imply that strong
> > authentication has been performed. IP checks are weak authentication
> > (and I would argue that they are not even authentic, given the number of
> > times I have seen that spoofed).
>
> I agree with you. However, the real problem is that system:authuser
> is getting overloaded. PT groups aren't flexible enough to handle the
> name space of users and machines. Not being able to put a large number
> of principals into one group, and not being able to put groups within
> groups, are the real problems. If those two were solved, then I bet
> reliance on system:authuser would drop.
>
> < Paul
>
Surely for our senario this would be a big win. We have reimplemented
NIS's (YP) Netgroups under Hesiod (DNS), to control who could log in to
what machine in the Realm and what machine could NFS mount what partition.
With AFS we have been scratching our heads wondering the best aproach
to granularizie AFS. I like Paul's method, as it gives me better
security to boot.
Randall
