From: hedrick@klinzhai.rutgers.edu (Charles Hedrick) Subject: Re: Your SLIP is showing... (I wish) Date: 16 Feb 1993 01:32:29 GMT
vera@fanaraaken.Stanford.EDU (James S. Vera) writes:
>One question: How does the traffic carried running txconn and X
>clients remotely compate with the traffic one would get running SLIP
>and running X clients remotely? Enquiring minds need to know!
I have to warn you that I've looked at the term code and documentation
briefly, but am far from an expert.
It doesn't look to me like there's a lot of difference, as long as you
are using V-J header compression. That reduces the size of headers
for SLIP to something on the same order as the headers that term uses.
If you aren't using header compression, then you'll be hit by the fact
that IP and TCP headers together are about 40 bytes, so there's that
overhead for each packet. As far as I can see, term simply routes the
X connection over one of its virtual links. It doesn't seem to be
doing the sort of X-specific hacks that X.remote uses to keep down
traffic. The same is true of SLIP. So I wouldn't expect much
difference between term and SLIP for X connections.
It looks like the primary advantages of term are (1) that you can run
it as a normal user, as long as you have access to machines at both
ends. With SLIP you normally need help from a system or network
administrator to set up the connection, and (2) that if your SLIP does
not do header compression, term is a big win.
SLIP's primary advantage appears to be direct access to the network,
which means you can talk to any service on any host without doing
advance setup, while term front-ends only some services by default.
(However it's possible to set up links for most services.) The KA9Q
implementation of SLIP that I maintain doesn't have a full set of
services, so term may be more flexible in some cases.
The fact that term uses a proxy on another system can be both good or
bad. It depends largely on how easy you find it to get access to a
system that supports SLIP versus how easy you find it to get access to
a lightly loaded Unix machine with good network connectivity.
On our campus (and many, I think), dialups are connected to terminal
servers. Those terminal servers support SLIP. When I run SLIP at
home, I'm connected through the terminal server directly to the
network. In order to use term, I would need to connect through the
terminal server to some Unix host and run term there. Thus all of my
traffic is being encapsulated twice, once in the term protocol, and
once in the TCP/IP connection between the terminal server and the
machine on which term is running. I am also subject to the load on
that machine.
This isn't a big deal for me. Our local network is fast, so I
wouldn't notice any delay due to the TCP/IP connection, and I have
access to lots of unloaded machines. For a typical student, things
could be more marginal. In many universities, students have accounts
only on systems that are fairly heavily loaded. If the machine
running term is loaded, this would introduce delay in their
communications. For other users it's even worse, because they don't
have access to Unix accounts at all. (I don't see anything about an
IBM mainframe implementation of term.) As a system and network
administrator, I'd rather see people using SLIP. It's one less
process running on our timesharing machines, and the overhead on the
terminal server is lower for a SLIP line than a telnet session. But
for many people, term is certainly a neat tool.
In summary, which one is better is likely to depend upon your specific
situation.