🚀 go-pugleaf

nagle@netcom.com (John Nagle) writes:

>mpiet@is.morgan.com (Mark Pietrasanta) writes:
>>Okay.  I'll ask it.  Is Client-Server going down the tubes in
>>favor of peer-to-peer?

>       The main purpose of client-server systems is isolation of the
>manipulation of important data from the (probably graphical) user
>interface.  Whether both reside on the same machine is not the central issue.
>GUI-based applications tend to be mostly user interface in terms
>of code volume, and breaking up the program into a user-friendly
>client and a server whose users are programs seems to be a good
>division of labor.


I think that client-server is the wave of the future.  The future that I
see is that the computer will be a network of machines.  Some of the machines
will be clients, some machines will be servers and some machines will be both.
The future I envision has a disk machine, which has a lot of online storage
connected to a fast but otherwise pretty minimal CPU (NFS).  There is a compute
server, a computer which has been optimized to have a fast, powerful computer
and lots and lots of RAM.  There are terminals connected to terminal servers,
terminals which are servers (X-terminals).  Security is maintained by a
security server (Kerberos).  Time is provided by a time server.  Now, the
machines can take on different roles at different microseconds.  For example,
the compute server is a client of the file server, but it is a server to
a general purpose computer.  The file server is obviously a server to lots of
clients, but less obvious is that it is a client of the security server and
the time server (Kerberos requires time synchronization).

I think if you look carefully at the new offerings, you will see that deep
down inside they are still client server based!  

Jeff Silverman, Boeing.