Speaking of the need for the server to listen for a connection,
in fact, the "connection" between the server and the clients is
"established" initially in the disklist file, which names the
client machines and the partitions to be backed up.

With the "connection" having been "established" in
the disklist file, the server "agrees" to contact the clients to
back them up.

So as long as one allows that the configuration can be part of the
"connection" between the server and the clients, there is no abuse of
language...

-----Original Message-----
From: Lengyel, Florian
Sent: Sun 12/4/2005 3:06 AM
To: amanda-users AT amanda DOT org; amanda-users AT amanda DOT org
Subject: RE: "client" and "server" terminology backwards in the docs?

Now you have a counterexample to the rule that servers must
passively listen for a connection and respond. Servers provide
a service; clients make use of it. The implementation
specific detail that servers have to passively listen for a
connection is unnecessary. It's "natural" (or habitual) to think
of the backup server as  the server; the client machines
are making use of the backup service. In this case, the amanda server
is bending over backwards by contacting the clients, so that they
don't have to take the trouble to contact it (or because they won't
on account of not knowing what's good for them).
Isn't that thoughtful of the server? The AMANDA server is being morally
good by following the moral ideal to remind the clients they should
be backed up, instead of merely adhering to the moral rule to do its
duty only if the client contacts it.


All right, one can think of the clients as agents. Perhaps that's
a better term than client. It's not original: Veritas calls its
"client" processes backup agents.

But, in the end, if the use of client instead of agent is a hurdle,
then one can put on one's resume that the terminology defeated
all attempts at comprehension.

-----Original Message-----
From: owner-amanda-users AT amanda DOT org on behalf of Jon LaBadie
Sent: Sun 12/4/2005 12:51 AM
To: amanda-users AT amanda DOT org
Subject: Re: "client" and "server" terminology backwards in the docs?

On Sat, Dec 03, 2005 at 10:36:36PM -0700, mindfuq AT verizon DOT net wrote:
> I've been reading the documentation on amanda.org, and it seems the
> authors have this terminology flipped.  Servers always passively
> *listen* for a connection, while clients are active initiators.  In
> the Amanda model, the centralized backup host is actually a *client*,
> because it's the active "consumer" that initiates connections, whereas
> all the nodes on the network that have data to backup are servers
> listening for a client - and serving the data on request.
>
> I find it confusing to read the Amanda documentation, because it
> appears the Amanda Core Team is calling the server a "client", and
> vice versa. 

You are trying for absolutes and there are precious few.

Another example of a "server" initiating the connection is
the X server when it initiates the connection with X terminals
or PC serving as X terminals using the XDMCP protocol.

--
Jon H. LaBadie                  jon AT jgcomp DOT com
 JG Computing
 4455 Province Line Road        (609) 252-0159
 Princeton, NJ  08540-4322      (609) 683-7220 (fax)