On 10 Jun 1999, Greg Stark wrote:
> Matt Kern <email@example.com> writes:
> > In line with normal KerberosV installations I am planning on installing
> > kerberised versions of rsh, rcp, rlogin, ftp and telnet in /usr/bin. The
> > kerberos packages will conflict with telnet and divert the netstd packages
> > to binary.netstd (which the kerberised versions will use as a fallback).
> > ...
> When I did the kerberos 4 packages I had them use a prefix of k. Actually I
> used a transform of s/^k?/k/ Ie, prefix a k if there isn't a k prefix already.
> This means they install as ktelnet, krsh, krcp, etc.
Many Kerberos IV installations do have the 'k' prefix, whereas I have yet
to see a Kerberos V installation that does the same.
> I am wary of using diversions because several packages might divert things
> like telnet, rlogin, etc. diversions work best if it's really a unique
> situation, not when there are multiple packages that provide the same thing.
I would like to (and may well) conflict with the telnet package; however,
I can't conflict with netstd without causing loads of problems. I will
actually need netstd on the system to provide the fallback rcp/rsh
> Ideally we would have you install k5telnet k5rlogin etc, I would install
> k4telnet k4rlogin, etc, and we would use alternatives to allow the user to
> choose which would be the default telnet and rlogin. But this requires
> cooperation from the netstd package.
This sounds like a good proposition -- It gives those with experience of
Kerberos the ability to swap in Kerberos IV/V functionality where they
expect to find it (telnet as opposed to ktelnet) without taking away
peoples options. What are your views Herbert?
Do we really want the two Krb4 and Krb5 packages on the same system? MIT
KerberosV will do both and my current setup works with both.
\\\\///// Matt Kern Tel: (01223) 355588
| | firstname.lastname@example.org http://xanadu.pet.cam.ac.uk/~mwk20/
| O O |
| L | If I had better tools, I could more effectively
| \__ | demonstrate my total incompetence.