[463] in SIPB-AFS-requests
machines in pts database?
daemon@ATHENA.MIT.EDU (qjb@ATHENA.MIT.EDU)
Tue May 21 11:52:51 1991
From: qjb@ATHENA.MIT.EDU
Date: Tue, 21 May 91 11:52:27 -0400
To: jon@MIT.EDU
Cc: jik@pit-manager.MIT.EDU, eichin@ATHENA.MIT.EDU, sipb-afsreq@ATHENA.MIT.EDU
In-Reply-To: Jon A. Rochlis's message of Tue, 14 May 91 12:59:48 EDT <9105141659.AA10544@delwin.MIT.EDU>
The reasons that people stated for doing this one way or another
all seem valid to me. The main reason that I have for using a
special key is that this allows the maintainer of a service to
move or replicate the service to other machines without needing
intervention of either kerberos or afs administrators. This is
a minor point, however, and could easily be outweighed by some
of the other concerns. I wasn't really saying that my way is
better or worse, just that that is what I would probably do in
this circumstance. I don't see the need to add more granularity
than necessary. For AFS-related services (or things that use
afs) which server the service is on seems irrelevant, which is
why we have one afs key for each cell rather than for each
server. This is not the case with something like remote login
which is why we have a separate rcmd key for each host. For
news, it seems that we really need only one key for the whole
service. This is how I see it anyway...
Sorry for the delay in response; I was out of town.
