Kerberos is certainly a barrier to adoption, and that's why I'm *not*
dropping the PAM support.  it works, it's well tested, and it's a perfectly
reasonalbe implementation.  That will remain in the code permanently, and
we'll just recommend a Best Practice of use krb5.

Password caching is as good as dead...

On Fri, Jun 11, 2010 at 10:00 AM, Steven Jenkins
<[email protected]>wrote:

> 2010/6/11 Phillip Moore <[email protected]>:
> >
> > Well, AD *is* Kerberos, but it's another matter entirely to manage the
> > configuration of Kerberos on the UNIX clients.  If someone doesn't have
> > support for it today, adding it is non-trivial.
> > OTOH, it isn't that hard, once you understand how the peices fit
> together.
> >  Certainly, being able to help clients leverage this is a valuable
> service
> > to offer.
>
> Yes, and a former employer (as well as other companies) offers such
> services.
>
> Looking at the experience of AFS, though, I think that requiring
> Kerberos has been a significant barrier to adoption. EFS should not
> require Kerberos in all cases, however -- using the same security
> model as the underlying network filesystem is very understandable to
> new adopters, so ...
>
> NFSv2/3 - use uids and -trust- the client
> NFSv4 - use whatever model (uids, GSSAPI..)
> AFS - whatever model (Kerberos right now, but others are in the works)
>
> And, yes, password caching is bad.  Trusting the client is arguably
> worse, but at least trusting the client is what a site is currently
> doing.
>
> Steven
> _______________________________________________
> EFS-dev mailing list
> [email protected]
> http://mailman.openefs.org/mailman/listinfo/efs-dev
>
_______________________________________________
EFS-dev mailing list
[email protected]
http://mailman.openefs.org/mailman/listinfo/efs-dev

Reply via email to