[David Johnson <dave@mom.computone.com> writes:]
>Seems that namer will need to handle deaths at least of servers which are
>across the network. Maybe this is indeed useful for local servers also
>to maintain consistency. It is also frustrating to find a failure which
>is not obvious. Since a remote user could kill a server (with
>appropriate permission) a local user would not know why an access is
>failing.
You appear to be confusing the issue of when namer is used. Once a
connection is set up, the namer is out of the loop, so his knowledge or lack
WRT server condition does not help a user who happens to have a service
mounted and is trying to use it. The user will see I/O errors on his
requests until he closes files, unmounts the mount point, and so forth.
As a distinct issue, namer would continue to hold an entry for the server's
port_name. When you tried to mount it, you'd get an error indicating that
the port wasn't valid. What I *thought* we were talking about was whether
it was very important to make the entry go away.
Andy
Received on Mon Nov 28 11:07:48 1994
This archive was generated by hypermail 2.1.8 : Thu Sep 22 2005 - 15:12:10 PDT